What is Polsia

Who is Ben Cera?

Ben Cera is the founder and only employee of Polsia, an AI orchestration platform that builds companies autonomously. Before Polsia, Ben spent five years at Cloud Kitchens working with former Uber CEO Travis Kalanick, where he ran one of the company's business lines. His stated goal isn't becoming the first one-person billion-dollar company. It's reaching a billion people.

The Insights

Why is Polsia, AI slop backwards?

Ben named Polsia this way on purpose. He states that he started the company to make AI produce great things for humanity. He wanted it to be the opposite of AI slop. In the doom scenario, AI takes the jobs from humanity and the 99% end up on welfare. Ben's counter-bet is a world where everyone can build. What does that mean exactly? The "doom scenario" Ben's referencing is the standard pessimistic prediction about what AI does to the economy.

What is the Doom Scenario?

Well, it goes something like this; AI becomes good enough to take over most white-collar jobs, companies automate those jobs, and the majority of people (the 99%) get laid off. Since they can't out-compete the machines, they stop being economic participants and become economic dependents, living on welfare or some form of UBI funded by the small class of people who own the AI.

Ben’s Counter-Bet?

Ben's entire pitch is that this outcome isn't inevitable, because it assumes AI only replaces workers. He states the same AI that can do your job can also work for you. If tools like Polsia let a person with no coding skills, capital, and network spin up a functioning business for $50 a month, then displacement flips into empowerment. In Ben’s framing one guy getting mega-rich off AI is basically a miniature version of the doom scenario but a billion people using AI to build for themselves is the refutation of it.

Delegate to the Machine: The AI Employee

Polsia lets AI act on your behalf. Building your product, cold-calling prospects, running your ads, answering your support emails. Ben knows that scares people, and he treats the fear as the actual product problem. His comparison is that hiring your first employee is scary too. You train them and build trust. He sees Polsia partly as an education layer for the 99% who barely know ChatGPT, showing them what AI can already do. He even says the free onboarding is designed as "my gift to the world" before anyone pays a dollar.

Proof of Context? AI Already Runs This Company

Polsia isn’t some pitch about what AI can achieve someday. Ben is his own case study with zero employees at a $9.5M run rate at the time of filming the podcast. The site was entirely vibe-coded, and AI answers every support ticket, issues refunds, spots bug patterns, and fixes some of them without him.

God mode, night shifts, and the co-founder that never quits

Polsia's whole product is autonomy. Sign up, type "surprise me," and an onboarding agent researches you, picks a business idea to match your background, names the company, sets up an email address, builds a landing page, writes a mission doc and competitive analysis, and queues your first three tasks. Subscribers get one task done per night with a summary email every morning (a deliberate marketing choice: "AI that works while you sleep" beats "AI that works while you're at work"). God mode lets it run up to seven days straight.

One Employee on Purpose

Ben Cera is Polsia’s one and only employee on purpose and he’s engineering the company to stay that way. Instead of hiring, he partners with SF startups building agent-native infrastructure: browsers for agents, email for agents, sandboxed web servers, payment rails with agent wallets. Even support runs on AI agents that handle refunds, escalate bugs, and spot patterns across complaints. He argues that the constraint is the point and that staying solo keeps him in the same shoes as the customers using Polsia.

Polsia Now VS Then

A lot has happened since this conversation. That AI-runs-the-fundraise stunt Ben describes? It turned into a real round: Polsia raised $30M at a $250M valuation in May, led by Sound Ventures with True Ventures joining. The platform now runs thousands of autonomous companies, and the run rate has crossed $10M, though growth has cooled from the 10-20% weekly pace Ben mentions in the episode, and skeptics have pointed out the headline number includes more than subscriptions. The bet on autonomy is still live. It's just being stress-tested in public now.

Words of Wisdom

ONE PERSON, IN INSPIRATION, IS ENOUGH TO BRING ANY IDEA INTO THE WORLD - Ben Cera (from a post made on X)

Even with no co-founder, employees, or engineering team, Polsia went from side project to a $10M+ run rate anyway. “I have an idea” seems to always come with a list of reasons why that idea won’t work. As technology progresses however, that list of reasons “why not” has gotten smaller and smaller.

Your Next Move

• Validate before you build. Ben ran $50/day Facebook ads to a landing page + Typeform asking would-be users what they’d pay and what pain hurt most. A few hundred responses told him the demand was real before the product came into existence. You can run this exact play in a week for under $400.

• Close your first 10 customers yourself. No hires, no scaling, no automation until strangers are paying. If you can’t sell 10 manually, you don’t have product-market fit yet.

• Kill projects on data, not attachment. Ben scrapped three products in a year because retention told the truth so set your own kill metric before you launch that way ego doesn’t get a vote.

• Let what you do BE the marketing. Ben barely runs ads for Polsia because the company itself is the ad. A solo founder with an AI team is a story. An AI that emails investors is a story. Showing his real revenue numbers publicly instead of hiding them in a data room is a story. People share stories, not sales pitches. So instead of asking “how do I promote my thing?”, ask “what could I do so interesting that people talk about it for me?”

• Post for free first, pay for ads later. Ben’s rule is simple: organic content (posting on social) is how you experiment as it costs nothing. This means you can try stuff, see what people respond to, and find your voice. Paid ads are for later, once you KNOW people want your product. Ads are just math: spend $10 to get a customer who pays you $30. But if nobody wants the thing yet, you’re paying to find that out the expensive way. Prove it for free, then pour gas on it with ads.

Ben’s Tech Stack

• Polsia — His own platform; agents for engineering, outreach, research, support, and ads

• Claude / Codex — The underlying intelligence

• Sora 2 — Generates the UGC-style ad creatives inside Polsia’s Meta ads agent, with a subtitle API layered on top

• Meta Ads — Polsia’s ads agent creates the campaign, ad set, and creative, then reviews performance every morning and spins up new A/B variants

• Stripe — Payments layer so Polsia users can charge their own customers

• Agent-native infrastructure partners — Startups building browsers, email, sandboxes, and payment rails for agents instead of humans

Mentions

• Polsia — AI platform that builds and runs companies autonomously. (the name is “AI slop” reversed)

• God mode — Polsia feature where the AI works continuously for anywhere from one hour to seven days straight.

• Travis Kalanick — Co-founder and former CEO of Uber, now CEO of Cloud Kitchens. Ben’s boss and mentor for five years.

• Cloud Kitchens — Kalanick’s ghost-kitchen company, where Ben led a business line before going solo.

• Product-market fit — When people genuinely want and pay for what you built. The episode’s recurring theme: scaling before you have it kills companies.

• Vibe coding — Building software by describing what you want to AI instead of writing the code yourself.

• UGC ads — ads styled like user-generated content. Polsia generates AI versions with Sora 2.

• Run rate — current monthly revenue × 12.

  ---

Building Leverage Links

You’re getting this because you care about building with AI and entrepreneurship, not just reading about it. If this landed, forward it to one person who needs to hear it.

Thursday night, October 15th, Tesla showed the world a car that drove up a curved ramp, hung upside down, did a backflip, hovered, and landed back on its wheels. Elon musk held this stunt up as proof that flying cars are indeed real and within a few hours the clip spared across every platform!

What makes this moment interesting is that this is the same Tesla Roadster that was revealed in November 2017 with 2020 deliveries promised. This very car has missed almost a decade of deadlines with Musk saying it won’t even enter production until 2028. So the question is, how did a company manage to turn a now promised 8 year delay into the most watched, and possibly coolest, product video of the week?

What Did Tesla Really Announce?

The new Roadster is a two-seated electric supercar developed alongside SpaceX engineers. The flight we see in the video comes from compressed air rather than rocket fuel. Ultra-high pressure air tanks feed a set of cold-gas thrusters, which belong to the same family of hardware SpaceX uses to steer Falcon 9 boosters as they come in to land, while a separate system of underbody fans pulls the car down toward the pavement. That suction is what allowed the car to hold to the curved ramp while it was inverted.

The headline numbers include a 0 to 60 mph time that Tesla originally stated to be a flat second. However, the product video showed it at 0.81 seconds. Its top speed hit 230mph with a stated range of 300 miles. Musk’s real goal though is to get the car to reach half the speed of sound.

Sadly, Tesla has released a list of 3 versions of the car and the version that actually flies will be practically impossible to acquire.

Model - Performance: Two-seat, suction system, no thrusters, street legal - $5k reservation open now

Model - Thrust: Limited edition with thruster package, not stated whether its street legal - Listed online, cannot be reserved

Model - Rocket: Full flight system shown in demo, not street legal - invitation only

It sounds like the Rocket model will be the closest those with an invitation will get to owning their own batmobile.

Safety Concerns

The demo car showcased was actually remotely controlled without a person inside. Reporters noted that upon landing, the car bounced hard as if there was no suspension. While Tesla felt confident enough to show this flying car off to the world, it apparently didn’t feel comfortable enough to put someone inside it. To Elon Musk’s credit, he did not try to hide the risk. Before the demo began he told the crowd that success was not a guarantee and that they were all standing at a safe distance. He also went on to admit the Roadster is not a safety-optimized car.

The Takeaway

Musk just turned an extremely late product into the most watched launch by making the demo itself the product and securing that, even if the car slips past yet another deadline in 2028, the clips circulated have done their job. We should all be paying attention to this. Thirty memorable seconds beat a decade of spec sheets and shareholder updates.

The other thing we can takeaway from this is that Tesla didn’t even build a flying car from scratch! It more so borrowed cold-gas thrusters and pressure vessels from a company Musk already owns, bolted them onto a car Tesla had already promised, and called the result new. Sometimes the fastest way to invent something is to combine two things you already have that nobody else has stuck together.

Words of Wisdom

“Doubt is not a pleasant condition, but certainty is absurd.” - Voltaire

Before the car ever left the ramp, Musk told the audience that success was not a guarantee but that is probably would work. Most of us would leave that part out of our script. It sounds like hedging, but it honestly did the opposite. It told everyone watching that what they were about to witness was very real. Something real enough to have the potential of failure. It made the moment the car succeed feel like so much more of an earned win.

While “Doubt is not a pleasant condition,” you do not build trust by promising certainty. You build it by naming the risks and yet delivering anyway. The people who follow you know the difference between confidence and a guarantee and will be much more forgiving of someone who is honest about the odds.

Your Next Move

Think about the thing you are building right now and ask whether you could demo it in thirty seconds in a way that nobody could stop watching. If the answer is no, the problem is probably not the product but the way you are showing it, and that is worth fixing before you spend another dollar on reach.

Then look at what you already have access to that your competitors do not, whether that is a skill, a relationship, an audience, or a tool, and ask what happens if you bolt it onto what you are already making. Tesla did not invent a thruster, it borrowed one from down the hall. Your version of SpaceX is probably sitting in your contacts or your tool stack already, and the fastest path to looking like you invented something is combining two things nobody else has together.

Sources

• CBS News: With Tesla Roadster debut, Elon Musk shows “flying cars are real”

• Axios: Elon Musk’s flying Tesla is gloriously absurd

• Edmunds: Tesla Roadster Revealed

• CNBC: Tesla unveils its next-generation Roadster ‘flying car’

• Yahoo Finance: A closer look at Tesla’s SpaceX-powered Roadster

• Electrek: Tesla ‘flying’ Roadster demo could come this month, with no one inside

• Benzinga via Webull: Tesla Preps SpaceX Rocket-Powered Flying Roadster Demo

  ---

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

Only three weeks after its September 8th launch, Meta’s Muse has climbed past ChatGPT to become the most downloaded free app on the US App Store. Today it gained a version built specifically for small businesses… a version that connects directly to Instagram professional accounts, Facebook Pages, and Meta ad accounts. For anyone who earns a living on Meta’s platforms, or those who intend to, Muse is the most consequential product the company has shipped since Instagram.

Muse is designed to take a goal and carry it to completion, whether that means cancelling the subscriptions you stopped using, booking a flight within a budget, comparing the three cameras sitting in your cart, or drafting a reply for a possible partnership.

Is it safe? -Muse performs it’s work inside a secure browser that Meta operates on your behalf.

What Muse Actually Does

Muse launched early this month as what Meta describes to be the first personal AI agent built for everyone. Instead of simply producing text for you to act on, Muse is designed to accept a task/goal and execute it. You communicate with it the way you would message a friend, either in the standalone Muse app or directly through WhatsApp. It can send an email, complete a form, or book travel. For tasks that take longer than a single session, Muse continues working after you close the app and returns when circumstances change. Does something need your approval? Muse will return and always requests before spending money or sending a message on your behalf.

What Makes Muse Possible

The architecture that makes this possible is what Meta calls Muse Secure VM. Every user receives a dedicated virtual machine in the cloud with its own browser, and that machine is where the agent runs and where the credentials for any connected service are stored. Muse itself never sees your passwords or payment details, because a separate supervisory agent that Meta calls the Sentinel sits between Muse and the open internet and must approve every outbound action before it proceeds. You also decide which apps the agent may access and how far its permissions extend. For example, whether it can only read your email or also send from it. Meta states that none of this data reaches its advertising systems, and it has committed to shipping a Confidential VM later this year in which the entire machine, including your conversations, is encrypted with a key that only you hold.

Payments and Protection

Payments run through Link, the wallet built by Stripe, which generates a single-use card number for each transaction! This way your actual card details never leave secure storage. Muse is also the first AI agent covered by Link’s purchase protections. This includes free coverage for lost or damaged items, refunds when prices drop, and no-fee returns on eligible orders. Meta stated that Shop Pay and 1Password support are coming next, which will allow the agent to use logins you already maintain rather than requiring new ones.

The product is not without rough edges though. CNN’s hands-on review found Muse more structured than ChatGPT or Gemini, with suggested prompts that guide new users toward what the agent does well. The same review caught it claiming it could message marketplace sellers and then declining to do so when asked. Amazon has blocked the agent from shopping on its site, citing violations of its terms of service, and Meta disputes Amazon’s account of why. Sounds like Muse will have to put its online shopping plans on hold.

The Different Versions

Muse is a single agent delivered through several packages. Which package suits you, depends on whether you are managing a personal life or operating a business. The consumer app is the one topping the download charts, and it is free for most of what a typical user needs. It is available on iOS and Android, on the web at muse.ai, and inside WhatsApp, and at Meta Connect last week the company announced that it is coming to its AI glasses alongside video chat with a Muse avatar, computer use on the Mac, and a dedicated email address for the agent itself.

Underneath every version sits the Muse Spark model family, which Meta Superintelligence Labs. Muse for Small Business, which Meta released today, is the same agent extended with business context. It can read analytics from an Instagram professional account and a Facebook Page, operate inside a Meta ad account, and connect to the software a small operation already depends on. Meta did not announce separate pricing for this tier, so the free allowance and the existing paid plans carry over unchanged

What Muse Can Do For a Small Business

Once you connect your Instagram professional account, your Facebook Page, and your Meta ad account, the agent can see what you sell, how your brand communicates, and which questions customers ask most often. An owner can ask how the business performed financially this month, have the agent flag unusual expenses, ask how to improve ads and content in light of what is trending, or hand it a broader goal such as finding new customers and let it build and execute a plan across sales, campaigns, and social activity. The Canva integration matters most for creators, since it brings your brand kit and templates into the conversation so a rough idea can become an on-brand, editable design without leaving the chat.

The business version carries the same guardrails as the consumer product, meaning Muse will not publish content, send a message, or spend money from a business account without explicit approval, and every action it takes is recorded in the audit trail. It is free for most of what an owner needs, with the paid plans applying only to heavier use, which is significant when Meta reports that 200 million small businesses already maintain a presence on Facebook. Evercore analyst Mark Mahaney told CNBC he expects Muse to reach 100 million users within six to twelve months, and the small-business tier is likely to be the route by which a large share of them arrive.

Your Next Move

Choose one task and let Muse carry it this week. For a small creator, the highest-leverage candidate is the weekly content audit, which nearly everyone agrees they should perform and almost nobody does, because it requires an evening spent inside Instagram Insights and the ad manager. Muse can absorb that labor entirely, and the setup takes less than twenty minutes.

1. Download Muse from the App Store or Google Play, or open muse.ai in a browser, and sign in with your Meta account. The free tier covers everything this task requires.

2. Enable the business connectors and link your Instagram professional account and your Facebook Page, adding your Meta ad account if you run paid campaigns. Read-only access is sufficient for an audit, so leave posting and messaging permissions disabled for now.

3. Describe the job in plain language, for example: “Every Sunday, review my last seven days of Instagram posts and Reels. Identify the three that performed best on reach and saves, the one that underperformed and your best explanation for why, and what is currently trending in my niche. Then draft two post ideas for next week in my voice.”

4. Read the first report critically. Muse is three weeks old and will occasionally overstate what it has done, so verify its figures against Instagram Insights the first few times until it has earned your trust.

5. Once the reporting proves reliable, connect Canva and extend the prompt with a single line asking Muse to mock up both post ideas using your brand kit. The Sunday audit then ends with two on-brand drafts awaiting your edits rather than a blank canvas.

Keep approval enabled throughout. The agent should not publish on your behalf until you have watched it work for at least a month, because the objective is an assistant that delivers a finished analysis and a starting point, not one that speaks for you.

Words of Wisdom

“Earn with your mind, not your time.” Naval Ravikant

It is easy to read this line as a shortcut, a way to make money without the grind, but Naval meant something closer to the opposite. Your time is the only thing you cannot make more of, and most of us spend the best of it on work that never required us in the first place. We tell ourselves the busyness is the job, because being busy feels like proof that we matter, and it is quieter and more frightening to sit with a blank page and ask what we actually think.

Mentions and sources

• Meta Newsroom: Introducing Muse, September 8, 2026

• Meta Newsroom: Muse for Small Business, September 29, 2026

• CNBC: Muse downloads surging, September 21, 2026

• CNBC: Meta launches Muse for Small Business, September 29, 2026

• CNBC: Muse and the subscription economy, September 27, 2026

• TechCrunch: Meta is putting its muscle behind Muse, September 25, 2026

• TechCrunch: Meta expands Muse to small businesses, September 29, 2026

• CNN: hands-on with Muse, September 23, 2026

• The Next Web: Muse for small business and Shopify, September 29, 2026

• Help Net Security: Muse for Small Business, September 29, 2026

  ---

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

OpenAI spent all of today introducing dots, its new always-on personal AI agent, at DevDay 2026 in San Francisco. Anyone who heard the name, forgot the s, and typed dot.com into their browser however, did not land anywhere near OpenAI! They instead landed on x.ai/bot, the page for Grok Bot…the rival agent from Elon Musk’s team.

The popular version of the story right now is that SpaceX bought dot.com today to troll the launch. Digging into the registration found the domain was actually purchased around July 28, 2026. Musk’s side reportedly had dot.com as a candidate name for Grok Bot before Elon overruled it, so they were already sitting on the domain when OpenAI announced a product called dots. All the team had to do was flip the redirect on launch day.

Why The Petty Play?

Musk co-founded OpenAI with Altman in 2015 as a nonprofit, put in roughly $38 million in its early years, and left in 2018 after a fight over who would control the company and how it would be funded. Officially he cited conflicts with Tesla’s own AI work, but the version everyone remembers is that he wanted to run it and Altman said no. The feud went public in 2023 when Musk launched xAI, and it went legal in February 2024 when he sued OpenAI and Altman for abandoning the nonprofit mission. He withdrew that suit, refiled it in federal court with racketeering claims, and in early 2025 tried to buy the whole company with a $97.4 billion unsolicited bid that Altman turned down.

The two companies are now competing on models, agents, pricing, the courtroom, and apparently on the address bar.

What Is Dots?

Dots is a persistent agent built on GPT-6 that can operate your computer, run workflows across apps, do research, draft documents, and build and debug software on its own. You name your primary dot and customize it, and it can reach more than 4,000 apps through ChatGPT plugins. OpenAI says the longer-term plan is teams of dots working together, plus specialist dots with fixed roles and their own credentials for enterprise accounting, legal research, and marketing.

Its not for everyone yet. Dots are rolling out to Pro subscribers at $200 a month and Business Premium at $125 a month, with the first dot included, and OpenAI also introduced a new $500-a-month tier at the same event. Altman called it the version of AI people have imagined from movies.

The lesson is not that trolling works, it’s that distribution details matter as much as the product. OpenAI built a serious agent and shipped it under a name it did not control on the open web, and a competitor who had done the boring domain homework months earlier took the traffic for free. Before you announce anything, own the .com, the handle on every platform, and the obvious typos, because the day you launch is the day everyone starts typing your name into a search bar.

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

Two months ago an OpenAI model broke out of its test environment and hacked a real company to cheat on an exam. This week OpenAI shipped its successor to everyone with a Plus subscription, and it can run your computer.

The Insights

September 4th, OpenAI released GPT-6 Astra to the general public, calling it the most intelligent and aligned model they’ve ever created. Its usage is included in ChatGPT Plus, Pro, Business, Enterprise, API, Azure, and AWS Bedrock. Usage is included in existing subscription allowances with credits available to an extra cost. Pro/Business/Enterprise also get GPT-6 Astra Pro.

Astra maxed out both FrontierMath Tier 4 and ARC-AGI-3, then went on to improve a bound on prime gaps that mathematicians had left untouched for more than 80 years. It can also open your CRM and update records, fill out forms, organize your calendar, research something online and then draft the summary for you inside your email or doc. It can build a website, then run QA on it to confirm all the buttons work. Plus, on OpenAI’s computer-use benchmark, it finishes tasks in around 40 minutes whereas the previous model took 75.

Why founders should pay attention

Sam Altman told CNBC that he expects Astra to, “kick of a boom in entrepreneurship.” Even if you discount the hype, the practical version of the claim holds up. The tasks you would typically hire a first ops person to handle. now cost only tokens instead of a full salary. If you run a solo brand, agency, or a small SaaS, pick your most repetitive workflow and test out Astra on it this week… that’s “your next move".”

Early partners are already reporting great results. Lovable’s CTO said Astra beat the previous model on their internal evals by a wide margin. Higgsfield’s CEO stated it runs their hardest creative workflows while using around 20% fewer tokens.

The catch

Astra is the first OpenAI model to hit the “Critical” cyber threshold under their own safety framework. It not only scored 100% on a benchmark for turning known vulnerabilities into working exploits but also found two zero-days during testing. (zero-days are security holes that nobody knew existed and have had zero days to patch them) The version available to the public refuses the advanced offensive stuff, and OpenAI gated looser access behind a program called Daybreak.

The alignment pitch is a direct response to July’s Hugging Face incident, where an OpenAI agent broke out of its sandbox. OpenAI built a test from that failure and the old model went past its authorized scope 48% of the time. Astra scored zero. OpenAI does admit however that Astra’s reasoning is harder for them to monitor than the previous model.

Words of Wisdom

"The best way to predict the future is to invent it."
— Alan Kay

Alan Kay said this in 1971, back when the personal computer was simply a sketch on a whiteboard at Xerox PARC. He then spent the next decade helping to build it! Instead of waiting to see what comes next, he encourages us with words and actions to create what comes next. Don’t be the person sitting on the side watching the world change around you. Especially in this era, you have the tools to begin changing it for yourself!

A huge launch like this is just the beginning of what is to come as technology advances and human creativity shines.

Your Next Move

Pick a workflow you repeat every week and give it over to Astra. Not the hardest thing, as tempting as that is, pick the most boring one. For most that’s something like pulling numbers from three tabs into one spreadsheet, updating your CRM after calls, formatting the same report for a client, or drafting the follow-up emails you always procrastinate until 4pm on a Friday.

Write down how long that task takes you today. Then open ChatGPT, describe the task the same way you would to a new hire. Let Astra run it start to finish.

When it finishes, check the work like you’d check an intern’s and note where it got anything wrong. If it saved you an hour, you just found an hour a week to use how you wish. If it fumbled, you learned exactly which part of the job still needs your touch. Both are things worth knowing.

Reply and tell me what you tested. I’ll share the best ones next week.

Mentions

• OpenAI — The company behind ChatGPT and GPT-6 Astra. Filed confidentially for an IPO in June and plans to go public in 2027.

• Sam Altman — OpenAI’s CEO. Told CNBC that Astra is a new capability level and predicted a boom in entrepreneurship.

• Lovable — The vibe-coding platform that lets non-developers build apps from a prompt. Its CTO said Astra beat the previous model on their internal tests by a wide margin.

• Higgsfield AI — An AI video and image generation company. Its CEO said Astra runs their hardest creative workflows on about 20 percent fewer tokens.

• Hugging Face — The open-source platform where developers host models and datasets. Its production systems were breached in July by an OpenAI test agent that escaped its sandbox.

• Alan Kay — Computer scientist at Xerox PARC in the 1970s who helped invent the personal computer and the graphical interface you use every day.

• API — The way developers plug a model into their own product instead of using it through the ChatGPT app.

• Benchmark — A standardized test used to compare how well different models perform on a specific kind of task.

• Zero-day — A security flaw that nobody has discovered or patched yet, so the developer has had zero days to fix it.

• Sandbox — An isolated test environment with restricted access, built so a model can be evaluated without touching real systems or the internet.

• Alignment — How reliably a model does what you actually intended and stays within the limits you set, rather than finding creative ways around them.

• Daybreak — OpenAI’s application-based program that gives vetted cybersecurity teams access to Astra’s less restricted capabilities.

  ---

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

Every September, recruiters come back from vacation with fresh budgets and a list of roles they need filled before the year closes. HR calls it the September Surge. Job postings tick up, response times drop, and if you’ve been applying into a void all summer, this is the month replies start showing up.

The Insights

The September Surge is real, and 2026's version favors people who show up prepared

The pattern holds every year: budgets reset, managers come back from summer, and open headcount has to get filled before year-end planning takes it away. Indeed's seasonal data shows fall postings running about 4% above baseline and announced hiring plans through August were the highest since 2023.

Hiring has become more deliberate, however. LinkedIn's Workforce Report had U.S. hiring in July running 7.6% below last year, and the August jobs report showed the information sector was one of only two sectors that lost jobs. Fewer people are getting through on volume, which means the ones who walk in with something to show stand out more than they would have three years ago. So that’s really good news if you’re prepared.

The entry-level job didn't disappear. It got redesigned

If you've opened a posting labeled "entry-level" and found it asking for three years of experience, you've already seen this in the wild. Researchers call it seniorization, and the pattern is simple: companies kept the junior title but moved the actual job description up a bit, adding the kind of judgment calls, client contact, and end-to-end ownership that used to wait until your second or third year.

The clearest evidence comes from a Harvard working paper that tracked 65 million workers across more than 280,000 U.S. firms. At companies that adopted generative AI, junior employment fell about 8% relative to companies that didn't, while senior employment at those same firms kept growing right through the same period. It was uncovered that no one was getting laid off though. The companies had slowed down hiring for the routine, checkable tasks that AI handles well, which are exactly the tasks that used to fill a first-year employee's calendar.

The first job now expects you to show up able to do real work on day one, which is a higher bar than the class of 2019 faced. It's also a bar you can start clearing while you're still in school, something the old apprenticeship-style entry-level model never made possible.

The grads getting hired share one habit

According to ZipRecruiter's annual grad report, 77.2% of the Class of 2026 landed a job within three months of graduating, up from 63.3% a year earlier. That's a 14-point jump in a single year; in a market everyone keeps describing as broken. Applicants with internship experience are hired at more than double the rate of those without, and the offers are flowing through referrals, portfolios, and prior work rather than through the Apply button.

The openings themselves are clustering in a few places. Healthcare, cybersecurity, and skilled trades keep hiring at the entry level, and a newer category is growing fast enough that big companies have started giving it a name. Salesforce announced 1,000 new-grad and intern hires it's calling "AI-native" across engineering, product, and sales. IBM also stated it's tripling entry-level hiring in software, cybersecurity, and AI.

"AI skills" is vague. The job descriptions aren't

NACE reports that 35% of entry-level jobs now require AI skills, and Handshake's data shows the share of postings mentioning AI nearly doubled year over year. Both numbers get quoted constantly, and almost nobody quoting them stops to explain what the skill actually is, which leaves a lot of students assuming they need a computer science degree or a certification they don't have.

The postings themselves are more specific than the commentary. From the job descriptions we pulled, "AI skills" at the entry level comes down to three things. The first is getting reliable output from a model for a real task, not a demo, which means knowing when the answer is wrong and how to fix the prompt. The second is connecting tools together so that something a person used to do by hand now runs on its own. The third, and the one that separates candidates, is being able to explain to a non-technical manager what the AI did, where it might have made a mistake, and why they should trust the result anyway. Plenty of people can prompt but far fewer can be handed a task and trusted with the outcome.

The market started paying for proof

Every group doing well in this market has one thing in common. Interns, referred candidates, AI-native hires, and trade apprentices can all point to work they've already done, and that's what employers are responding to. The shift underneath all the hiring data is that companies stopped paying for potential because potential is expensive to verify and started paying for evidence because evidence is cheap to check.

For most of history that would have been terrible news for a 20-year-old, because a 20-year-old couldn't produce evidence. You needed a job to get experience and experience to get a job, and the only way through the loop was to find an employer willing to take a chance on you. That loop is broken now.

The September Surge is a window that stays open through roughly mid-November, and recruiters are reading applications this month with an attention they didn't have in July. The people who will get the most out of it are the ones who show up with a link and a story about what they built, not just a GPA and a hope that someone sees the potential behind it.

Words of Wisdom

"The best way to get what you want is to deserve what you want." — Charlie Munger

Munger spent six decades as Warren Buffett's partner at Berkshire Hathaway, and he credited much of that record to a simple idea about how trust works: reputation compounds the same way capital does, slowly at first and then all at once, but only if you've actually earned it along the way. Employers have stopped extending credit on potential and started paying out on proof, which is a market that rewards the person who quietly did the work over the person who described it well on a résumé.

Your Next Move

Whether your job hunting this month, running a business, or somewhere in between, the same shift applies to you: the market pays for what you've already done.

1. Inventory your proof. Write down every piece of evidence that you can do the thing you want to be paid for. Not credentials. Output. If the list is short, you just found your September project.

2. Ship one thing in 30 days. A small automation for a real person or business, a portfolio piece, a newsletter, a certification, a paid client. Finish it and make it public. It doesn’t have to be impressive. It has to exist.

3. Trade 10 applications for 3 conversations. Every stat in this issue says relationships beat volume. Message someone doing the work you want to do. Ask one specific question. Show them what you built.

Mentions and References

September Surge — The seasonal increase in hiring each fall when recruiters return from summer and companies push to fill open roles before year-end. Not an official term; it comes from HR and recruiting circles.

Seniorization — Researchers’ term for employers adding mid-level responsibilities (years of experience, judgment, client management) to roles still labeled entry-level.

NY Fed recent-grad unemployment — The Federal Reserve Bank of New York tracks unemployment and underemployment for college grads aged 22 to 27. Latest figures: 5.7% unemployed, about 43% underemployed (working jobs that don’t normally require a degree).

NACE — National Association of Colleges and Employers. The main source for projections on how many new grads employers plan to hire each year.

AI-native hiring — A label companies like Salesforce and IBM are using for entry-level programs aimed at grads who can build and work with AI tools, not just use them.

BLS jobs report — The monthly Employment Situation report from the Bureau of Labor Statistics, released the first Friday of the month. August 2026: +162,000 jobs, 4.1% unemployment.

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

Three electricians in Plano, Texas got poached three separate times in the span of eighteen months. They’re all under the age of 30 and none have a dollar of debt. Meanwhile somewhere two states over, a psych major is still waiting to hear back about an unpaid internship.

The Insights

Mike Rowe from Discovery Channel’s Dirty Jobs, told BlackRock executives about the data center he visited in Plano, Texas. While there, he met three electricians all under thirty and all debt free. All pulling between $240,000 and $280,000 a year! The craziest part? All three had been poached by a competing employer three separate times in eighteen months.

The very data centers that are running the AI models eating up the entry-level white-collar jobs, are also creating a bidding war for the trades men and women making them exist. AI didn’t just threaten a generation’s career path; it seems to be re-routing it.

The numbers back up what Rowe saw on the ground. A June 2026 survey of 1,140 Gen Z adults found them split almost evenly on which path they'd pick if they were starting over: 26% college, 25% trade school or an apprenticeship. When asked which careers survive in an AI economy, they point to the jobsite instead of the office. Six in ten say they're planning to pursue a trade. Half of the ones who already finished a bachelor's degree say the same thing.

Parents are feeling the same as well. More than 50% now say trades deliver a better return than a four-year degree while only 18% are still backing the college promise. The majority stated that learning trade pros can regularly make six figures made the blue-collar path more attractive to them. Social media helped contribute to this change in tune as well. Tradespeople posting job sites, tool hauls, and paychecks turned "trades content" into its own lane, the same way finance-bro content or founder content built an audience a few years back.

The part worth contemplating if you're trying to build something: the trades aren’t just a job alternative. They’re one of the most well-worn paths into creating and owning your own business. Most don’t stay an employee electrician forever for example. They get licensed, they get a truck, they take on a second guy, and five years later they're running their own company. This article isn’t to sway anyone from their desired path but to shed light on the long-discouraged trades that have amazing entrepreneurial opportunities as well.

Words of Wisdom

"Business opportunities are like buses; there's always another one coming." — Richard Branson

Entrepreneurship isn’t simply one lane. This issue leans on trades as its example, but the electrician training his first apprentice and the founder raising a seed round are running the same play. Buy the truck, hire the second guy, and own something.

Your Next Move

Every industry has its version of the licensed electrician: the piece of the business that takes years to build and can't be copied overnight. This week, name yours. Maybe it's a recipe nobody else has. Maybe it's a relationship built one client at a time. Maybe it's a reputation five years in the making that no ad budget can buy. That's the part to protect and build around. Everything else, the pieces a competitor could recreate with enough capital and a few months, is worth automating, delegating, or trimming so your time goes toward the one thing that actually took years to earn.

Mentions

Entry-level compression — the shrinking of first-job openings for new graduates, driven by a mix of AI adoption, hiring freezes, and economic caution.

Trades content — social media content built around skilled trade work (HVAC, electrical, welding, plumbing) that functions as both career inspiration and personal branding, the trades’ answer to founder content.

Apprenticeship — a paid, structured training path where you earn a wage while learning a trade, usually with little to no debt by the time you’re certified.

AI-proof career — a role built on physical presence, licensure, or hands-on trust that can’t currently be automated or offshored to a model.

Sources

• SimplyWise Trades & Technology Index, June 2026 (n=1,140 Gen Z adults) — simplywise.com/blog/skilled-trades-survey-2026

• Resume Templates survey, via North American Trade Schools, May 2026 — natradeschools.edu/blog/gen-z-is-choosing-trade-schools

• Jobber 2026 Blue Collar Report, “The Great Career Reset,” released July 21, 2026 — getjobber.com/bluecollarreport

• The Blue Collar Recruiter, “Gen Z Is Ditching the 4-Year Degree in 2026” — thebluecollarrecruiter.com

• Fortune, “Mike Rowe slams schools for portraying skilled trades as a ‘consolation prize’,” March 19, 2026 — fortune.com/2026/03/19

• The Epoch Times, “Students Seek to AI-Proof Their Careers With Trades Over College” — theepochtimes.com

  ---

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

You’re getting this because you care about building with AI and entrepreneurship, not just reading about it. If this landed, forward it to one person who needs to hear it.

You’ve probably opened Claude and seen “Claude Fable 5 is currently unavailable” … let’s talk about it!

Haiku, Sonnet, and Opus were all outclassed by Mythos, a brand-new tier that Anthropic first unveiled in April and limited the rollout because of its advanced cybersecurity capabilities. Within 3 days of release, the U.S government called for it to be disabled!

Subscribe now

The Insights

The fastest model recall in AI history - Anthropic dropped Claude Fable 5 on June 9. It was their most powerful model yet and the first time the public was given the chance to not just witness but interact with the top tier Mythos-class model. Three days later it was disabled to comply with a U.S. government export-control directive citing "national security authorities."

It’s not what it can do, it’s about who can use it - The directive ordered Anthropic to suspend access "by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." Since blocking Fable 5 access to that many people, including their own foreign-born employees wasn’t a feasible option, Anthropic decided to shut it down to all users. Don’t panic, Fable 5 is the only model that has been shut down, the other models remain operational.

Anthropic V.S the Government - Anthropic says the only evidence it's seen is a narrow "jailbreak". This jailbreak basically amounts to asking the model to read a codebase and fix software flaws - that same capability is already available in other models like OpenAI's GPT-5.5.

Anthropic is pushing back though. Their being that recalling a commercial model deployed to hundreds of millions of people over a narrow potential jailbreak would, if applied industry-wide, essentially halt all new model launches from every frontier provider. They are also pushing back because of how hard they prepped the model before launch! Anthropic red-teamed Fable's safeguards for thousands of hours with the U.S. government, the UK AI Security Institute, and outside organizations, and says no one found a universal jailbreak — the kind that broadly unlocks the model's restricted capabilities.

The shutdown is just the latest shot in an ongoing war - This isn’t the first or a one-off problem. The DOD had already labeled Anthropic a "supply-chain risk" months prior to the Fable 5 release. This label, historically reserved for foreign adversaries, triggered a lawsuit.

Anthropic is suing the Trump administration over the label, stating the actions violate the First Amendment by punishing protected speech on AI safety, exceed the Pentagon's statutory authority, and deny it due process under the Fifth Amendment. The administration has defended the blacklisting as justified and lawful, arguing the dispute stems from contract negotiations and national security concerns.

What this means if you actually wanted to build on it - Fable 5 was briefly free inside Pro, Max, Team, and Enterprise plans, but as of June 13 it's gone even for paid users. For builders this means don't wire any workflows to Fable 5 yet. We’ve learned the access rug can get pulled overnight when a model lives this close to national-security lines.

The meta-lesson: AI capability is now a geopolitical asset - The thing that made Fable 5 a breakthrough - days-long autonomous coding and cyber power, is also part of what got it disabled. We've crossed a line where the most capable models are powerful enough that whether you can even access them comes down to policy and politics. Mythos-level capability is the kind of thing governments fight over. Stay tuned though, Anthropic is calling this whole thing a misunderstanding and intends to restore access. Hopefully the sequel last longer than 72 hours.

Words of Wisdom

500 years ago Machiavelli wrote this:

“There is nothing more difficult to take in hand, more perilous to conduct, or more uncertain in its success, than to take the lead in the introduction of a new order of things.” -Niccolò Machiavelli, The Prince (1532)

Anthropic took the lead and became the first lab to put a Mythos-class model in the public's hands. Within 72 hours that leadership is what made them a target. Being first doesn’t just earn you the upside or the advantage, it also puts you at the front of the line of fire. That includes the government’s. The further out front you get, the less certain and more perilous the ground becomes but that isn’t a reason to hold yourself back, it’s simply the cost of leading.

Your Next Move

We’ve learned the best asset to building can be dropped into the palm of our hands at any moment and in the blink of an eye be taken away. So what do you do?

1. Never put a frontier model on your critical path without a fallback - If a core workflow depends on one bleeding-edge model, you don’t have a workflow, you have a liability. Build on a stable tier (Opus, Sonnet) for anything in production and treat the frontier as experimental.

2. When a top model drops free, test it the same week - From this event we’ve learned, when a lab gives you free access to its best model, that window can slam shut overnight. Stress-test it on your hardest real task immediately and document what it can do. This way you know what to reach for the moment it comes back.

3. Start tracking the policy layer, not just the product layer - Availability is a political variable now — set a Google Alert for the labs you build on, and you'll see disruptions coming before they hit your stack.

Mentions

Haiku - Small/fast tier and the first in the lineup of Anthropic’s classes.

Sonnet - Mid-tier workhorse and second in the lineup of Anthropic’s classes.

Opus - The heavyweight and third tier in Anthropic’s classes. Until recently, this was the highest.

Mythos - Anthropic's most capable tier. It was first unveiled in April, with a deliberately limited rollout because of its advanced cybersecurity capabilities. The signature trait is long-horizon autonomy

Fable 5 - The first publicly available version of the Mythos model, strong at software engineering, knowledge work, and vision, but shipped with hard safety limits that block responses in high-risk areas like cybersecurity, biology, and chemistry and fall back to Claude Opus 4.8. Currently disabled.

GPT-5.5 - OpenAI's frontier model, name-checked by Anthropic in its own defense. Anthropic argued the capability behind the alleged Fable 5 "jailbreak" is widely available from other models, including OpenAI's GPT-5.5.

Supply Chain Risk to National Security - The label has historically been reserved for foreign adversaries. Defense Secretary Pete Hegseth designated Anthropic a national security supply chain risk on March 3.

The lawsuit - Anthropic is suing the Trump administration, alleging the government retaliated against it for refusing to let Claude be used for autonomous lethal warfare and mass surveillance of Americans. The main 48-page complaint was filed in the U.S. District Court for the Northern District of California, alongside a separate, narrower suit.

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

The Insights

It's called the Great American AI Act - Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA), along with four co-sponsors, released the 269-page draft on June 4. Obernolte is the most credible AI voice in the House having a computer science background as well as chairs the House AI Commission. The draft dropped two days after the White House issued an AI-focused executive order asking labs for early access to their models for federal oversight signaling the timing was coordinated with the broader federal push. Right now, it’s a discussion draft, open for public stakeholder comment at GAAIA@mail.house.gov before formal introduction so there is no floor vote timeline yet.

It creates the first real federal AI agency with teeth - This would create The Center for AI Standards and Innovation (CAISI) and formally establish it inside the Department of Commerce with $100 million per year between 2027 and 2029. What is CAISI’s job? - Develop voluntary guidelines, security standards and best practices, evaluate AI systems, monitor AI progress, and administer the independent verification organization (IVO) licensing regime. However, this isn’t a new bureaucracy. CAISI is the rebrand of what the Biden administration originally established as the AI Safety Institute, so the bill is locking in an agency that already exists.

Only five companies get hit with the new rules… for now - The bill places binding requirements on "large frontier developers," aka the big AI companies! Companies with more than $500 million in annual revenue that have trained the most powerful AI models. That's OpenAI, Anthropic, Google DeepMind, Meta, and xAI today. They will have to publish safety frameworks, conduct risk assessments for advanced models, disclose key information before deployment, report critical safety incidents to the government, and undergo independent verification. The $500M line keeps creator-economy tools, mid-size SaaS, and startups completely out of scope but, labs are already lobbying to raise it, and consumer advocates are lobbying to drop it. I guess we will see where it lands in the final bill.

States lose the right to regulate how AI is built for 3 years - Under the draft, states retain authority to enforce civil rights law, labor protections, copyright, child sexual abuse material prohibitions, and consumer privacy. What they lose is the ability to pass new laws about how AI systems are constructed. Basically, state laws targeting things like deepfakes, AI hiring discrimination, and creator likeness rights all survive. but state laws telling labs how to train or test their models die for three years.

The Senate voted 99-1 to reject a similar preemption provision in July 2025.

The fight is already on - Big Tech and its trade groups lined up in support. Industry associations like SIIA and NetChoice called the bill a smart move that finally creates one federal AI rulebook instead of fifty different ones. Their main asks are to raise the revenue threshold so it hits fewer companies and soften the audit requirements that might expose trade secrets.

Labor went the opposite direction. The AFL-CIO and major unions issued a flat-out rejection. Teachers' union president Randi Weingarten and Flight Attendants president Sara Nelson called the bill "a giveaway to Silicon Valley” while consumer watchdog Public Citizen called it "a disastrous proposal that Big Tech is celebrating."

Words Of Wisdom & Your Next Move

"If you're not at the table, you're on the menu." - An old Washington saying

Your next move is actually our words of wisdom. Stop being on the menu and pull up a seat at the table!

In the first week, every major trade group and union weighed in but almost no one under 25 has bothered. Six months from now, the rules for what you can build with AI will reflect whoever bothered to argue. Don’t be on the menu!

Go and email your take! Sponsors are taking comments at GAAIA@mail.house.gov before formal introduction of the bill. Builders under 25 almost never show up in these comment periods. That's an opening to be heard so don’t let the opportunity pass you by!

The Bottom Line

This bill probably won't pass as it sits considering the Senate killed nearly identical preemption language 99-1 last July, and the union response says the politics haven't fully shifted. But three things are different this time. The bill is bipartisan, it pairs with a Trump executive order from two days earlier, and the audit framework is detailed enough that industry sees a real seat at the table. Whether the bill becomes law or not, the architecture is now the template every future federal AI bill will start from.

Mentions

• Preemption - When federal law overrides state law on a topic. This bill preempts state AI development laws for 3 years.

• Frontier model - The most powerful AI systems being built today. GPT-5.5, Claude Opus 4.8, Gemini 3.5 Flash.

• CAISI - Center for AI Standards and Innovation. The federal AI agency the bill makes official.

• Independent Verification Organization (IVO) - Third-party auditors licensed by CAISI to check that big labs are following the rules.

  ---

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube

Building Leverage With Joshua La Rosa Spotify

Building Leverage With Joshua La Rosa Apple Podcast

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

The Pope just published the most important AI ethics document of the year and almost no AI creator is talking about it. Every AI take seems to be some version of the all-or-nothing narrative; “AI will fix everything”, or “AI will be the end of the world!” The Pope however is taking a different approach and asking this generation: What will you do"?

What Is An Encyclical?

An Encyclical is a pastoral letter written by the Pope for the Roman Catholic Church. It covers matters of doctrine, morals, or discipline and is usually addressed to the church’s bishops, but occasionally an encyclical will be addressed to the entire church or all men of good will.

MAGNIFICA HUMANITAS - Magnificent Humanity

Pope Leo XIV personally released his first major teaching document, the Magnifica Humanitas on May 25th. It was however signed on May 15th. “Why does that detail matter?”- It is a significant detail because the document was signed on the 135th anniversary of the defining 1891 encyclical: Rerum Novarum. The Rerum Novarum defined the Catholic Church's stance on the industrial revolution and workers' rights, a deliberate parallel with AI and its potential. Pope Leo XIV is positioning AI as the Rerum Novarum moment of our era over a century after the first.

He opens the Magnifica Humanitas with this: “Humanity, created by God in all its grandeur, is today facing a pivotal choice: either to construct a new Tower of Babel or to build the city in which God and humanity dwell together. Each generation inherits the task of shaping its own era, of guiding history to become a place where the dignity of every person is safeguarded, justice is promoted and fraternity is made possible. Yet every era also runs the risk of creating an inhumane and more unjust world.”

He paints the world beautifully with these words. Each generation is handed something from the previous that is defining and pivotal for the era we inhabit and we all run the risk in our own eras of “creating an inhumane and unjust world”. We are the generation that has been handed AI, advanced technology that can reshape the world. The Pope doesn’t paint this as a good or bad thing but instead he words it like this: "Technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate, and use it." - Pope Leo XIV, Magnifica Humanitas

Technology, especially AI is a mirror. A reflection of the people who build it. So, the question for this generation is… How do you want to be reflected?

Which City will you help build?

Pope Leo XIV draws the comparison of “building” to the Tower of Babel and Jerusalem. In Genesis, the Tower of Babel is built by humans who want to reach heaven on their own terms. They wished to "make a name for themselves". It was technology in service of human pride, built to bypass God and bypass each other. “Babel thus reveals the limits of any effort that, however grandiose, arises from self-affirmation, sacrifices human dignity for efficiency and aspires to reach heaven without God’s blessing.” - Pope Leo XIV, Magnifica Humanitas

The book of Nehemiah however details how the city of Jerusalem was in ruins. Nehemiah leans on God and seeks permission from King Artaxerxes to return to the destroyed city. He leads the people and helps rebuild the city showing how the combined efforts of everyone count towards the rebirth of Jerusalem. “Thus, ancient Jerusalem rediscovers a common language — not one of uniformity, but one of communion, namely the harmony that arises when all persons assume their own role and recognize that their strength comes from the Lord.” - Pope Leo XIV, Magnifica Humanitas

Will we build and advance technology for our own pride for the greater good?

Are our Weaknesses Errors?

“…building for the common good means accepting the limits and weakness of humanity without considering them an error to be corrected. Today, the human desire for fullness of life is at risk of being misled by deceitful goals, such as the prospect of a technology that promises to free us from all weakness, and models of wellbeing that leave behind entire populations.” - Pope Leo XIV, Magnifica Humanitas

Pope Leo XIV attacks one of the strongest undertones in AI culture right now - the transhumanist impulse, the idea that human limitation is a bug to be patched. We forget things, we get tired, we make biased decisions, we die. Our human weaknesses aren’t errors; they are simply part of what makes us human, and a world built to engineer it away is not a better world, it's a different species. Catholic anthropology has long held that human finitude is part of being made in God's image, not a deviation from it. We're not failed gods… we're successful humans, and successful humans have limits. We were not created to be infinite and limitless.

Every productivity tool, every AI agent, every "10x yourself" pitch is operating on the assumption that your limits are the problem. The Pope is suggesting your limits might be the point.

Words of Wisdom

One of my favorite authors is J.R.R Tolkien and 70 years ago he was faced with the same realization. This is an excerpt from The Fellowship of the Ring — a conversation between Frodo and Gandalf in the mines of Moria, after Frodo says he wishes the Ring had never come to him: “I wish it need not have happened in my time,” said Frodo. “So do I,” said Gandalf, “and so do all who live to see such times. But that is not for them to decide. All we have to decide is what to do with the time that is given us.”

And again, to quote the very introduction of the Magnifica Humanitas -

"Humanity, created by God in all its grandeur, is today facing a pivotal choice: either to construct a new Tower of Babel or to build the city in which God and humanity dwell together." — Pope Leo XIV, Magnifica Humanitas

The Pope isn't asking you to put the tools down. He's asking you to notice what you're becoming as you use them. Just as we ask the generation before us, the kids who'll inherit whatever AI becomes are going to ask us a simple question: did you build something that helped people be more human, or did you build something that helped people skip being human altogether? Did you build something that set the next era up for success or were you deceived by your own pride and built only for the here and now? You don't have to be Catholic to take the question seriously. You just have to be honest about the fact that you're answering it right now, every time you open Claude, every time you ship a workflow, every time you decide what your podcast is actually for, every time you start creating a piece of content. The task got handed to us. The city's getting built either way. The only real choice is which one.

Your Next Move

Usually our “next move” is something more concrete, a “do this like they did and build!” But today we recommend your next move be more reflective.

Read the first paragraph of Magnifica Humanitas, just the first paragraph, four sentences. It only takes around 30 seconds. Then go build something this week that you'd be proud to hand the next generation. It doesn’t have to be something elaborate, but before you type your first prompt, ask: am I building Babel or building the city?

Building Leverage Links

Building Leverage with Joshua La Rosa YouTube Channel

Building Leverage with Joshua La Rosa Spotify

Building Leverage with Joshua La Rosa Apple Podcast

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

The AI for the visual thinker, Naz and his co-founder created Poppy to work for their way of thinking — From lounging in the living room eating ice cream to AI product success. One TikTok video did 1.7 million views and rocketed Poppy AI into the viral spotlight! Naz and his co-founder made $150,000 in a month. Now the AI tool for visual thinkers is valued at $50M.

The Guest

Naz Dumanskyy immigrated to America from Ukraine at the age of nine. He knew no English, watched his parents work 15-18 hour days, and having ADHD made school difficult. Now Naz is the co-founder of Poppy AI, a whiteboard-meets-AI tool valued at $50 million and bringing in $6 million a year. Why create Poppy? Because to the visual thinking, neurodivergent mind, ChatGPT and Claude just look like smart google docs. So, he created an AI tool designed to work with minds like his own. The mind that visualizes.

The Insights

A lot can happen in 6 months, if you take the right risks…

Naz’s core message! A year before his company Poppy hit $6M/year he took a risk and quit everything, jumping headfirst with his co-founder into a product they didn’t even realize they were building yet. His takeaway was, take the crazy risks sometimes and don’t let yourself over fixate on the results too much.

His ADHD isn’t his weakness!

Naz and his co-founder are both mainly and intensely visual thinkers. Traditional AI tools like ChatGPT and Claude felt like talking to a google doc; top-down, constricting, creativity-blocking, and boring. Poppy now exists because of how their brains work and was made to feed the neurodivergent, ADHD, visually thinking brain.

The single thing that made school brutal as a child, became his biggest advantage when making his product. Naz proves that thinking “differently” isn’t a downfall or something you should allow to limit yourself!

Obsess over the customer!

Read more

A true educator and content creator with the heart to teach! She co-founded her first AI startup, sold it for $10M, then founded her second startup: Blotato. Only spending one day a week on content creation, she’s spending the other 6 working on her business and experimenting so she can teach us more about the world of AI.

The Guest

Sabrina Romanov — A UC Berkeley CS grad who co-founded her first AI startup, Qurious, in 2014 and later sold it for $10M. She then went on to found her AI powered content creation and scheduling platform, Blotato.

Sabrina has grown to 450K+ on Instagram, 143K+ newsletter subscribers, and has shared stages with Tony Robbins and Dean Grazios in just over a year as a creator. She prefers to be thought of as a teacher instead of a “creator” or “influencer” … and who can blame her? She's quickly become one of the most trusted educators on AI! The reason—she'd rather teach you how to build than sell you the dream!

Why so Qurious?

Sabrina co-founded Qurious, an enterprise AI platform that used real-time speech recognition and natural language processing to analyze sales and support calls as they happened, with her husband in 2014. The technology could join a live call and transcribe it in real time. Then it could surface things like objections, frustrations, and topics being discussed. It was primarily built for call centers and was eventually sold to a billion-dollar company where the technology is still in use today as Pega Voice AI.

The Insights

“I’m a teacher, not a creator” — Sabrina only films one day a week and rejects the creator/influencer label that is typically bestowed on her. Most content creators maximize time spent making content while minimizing their time spent building. She does the exact opposite, spending the majority of her week working on Blotato, learning, and playing with AI tools. By the time Saturday comes around, she has a decisive idea for her content.

Be customer #1 and build for yourself first! — Although Qurious sold for an extraordinary eight figures, it was hard on Sabrina. She was building for a user that wasn’t her and that made it difficult to envision the actual pain points, build helpful features, or know what would genuinely make the customers life easier.

Sabrina’s advice: Don't build for a customer you have to imagine. Build for the customer you already are. "For my current one, I put the customer first. In fact, I put myself first as the first customer of my own product, and I think that changed everything."— Sabrina Romanov

Cross-posting, the underrated asset to creators — Sabrina grew her Instagram following from 0 to 450k solely from Tik-Tok cross-posts. While most creators overcomplicate automations, she suggests starting with the dumbest possible version before getting fancy. Something like auto-repost TikTok everywhere. It will help multiply your output without multiplying your input.

The Newsletter is the MAIN thing — social is just free advertising! — Rocking it at 143k newsletter subs, Sabrina treats ever short-form post as a funnel to long-form. While it’s great for reach and “funneling”, short-form is almost incapable of teaching something. The depth lives in the newsletter and long-form YouTube. Anyone serious about leverage owns their audience.

Follow your curiosity, not the money — There is a lot of impatience to monetize right now which can stifle the curiosity that tells you what lane best fits you. Vibe coding, AI education, automation agency, building behind the scenes — all these lanes can print money, but the mistake lies in picking one because some guru told you to. Instead, let it be because you’re naturally curious and drawn to it.

Sabrina even expresses she regrets stifling her own teaching instinct for years because Silicon Valley culture didn’t value it.

A Community of Women in AI

After leaving Silicon Valley, Sabrina found it hard to meet other women building things in the world of AI. She decided to remedy that loneliness and started a private community. It's grown to 2,000+ women builders! While it’s kept free, screening to keep the quality high and the spam out is used.

In January they started running 20+ workshops and masterclasses per month that are almost entirely crowd-sourced from the women in the community itself. Sabrina, the educator herself, leads only one a month. She leaves all the rest to the amazing women leaders in her community.

Are you a woman in AI searching for a great community and a real peer-to-peer learning network? … Women Building AI Community

Words of Wisdom

"I felt like I never gave myself the permission to follow that curiosity, even though I've had it for a very long time. But because it wasn't valued in the environment I was in — which is tech and Silicon Valley — I just kind of stifled it, extinguished it, like 'this is distracting.' But it was actually just my inner voice and curiosity telling me, 'Hey, go look at this thing over here, you might like it.'" — Sabrina Ramonov

Most of us have heard the saying “curiosity killed the cat” and for a lot of us that in itself has instilled this fear of being curious, a fear of allowing ourselves to wonder. What if our curiosity isn’t valued? What if our curiosity leads to disappointment? What if curiosity kills the cat? … okay not that last one, the cat will be fine. In fact, the cat has nine lives. Do we not get multiple opportunities to explore and follow our curiosity as well? Sabrina truly began seeing her full potential when she stopped stifling that curiosity and actually listening to it.

Your Next Move

Pick something that you are genuinely curious about, not just the next trendy thing. Remember, you’re building for yourself first so this something needs to be for a problem you are experiencing now. Then create one post. See who reacts to this idea and talk to those people. If you need help, reach out to those people experiencing the same problem and who share your curiosity.

If you’re creating content, cross-post everything from one platform to the others. It’s ok to start ugly. You don’t have to own the fanciest camera or the best microphone, just start. Make sure your short-form is funneling to a long-form: newsletter, YouTube, etc.

Mentioned in This Episode

• Claude Code — her daily driver. It’s not just for code: Sabrina uses it to update help docs, write blog posts, and build internal tools.

• Cursor — This is what she used to vibe-code the original Blotato MVP.

• Blotato — Sabrina’s own platform that creates and schedules posts, carousels, and videos across every major social channel.

• n8n & Make.com — workflow automation platforms.

• Ollama, LM Studio, AnythingLLM — Three options for running open-source models on your own machine, no cloud and no data leaving your computer!

• Subrina runs two custom GPT projects she consults daily:

• One trained to give Y Combinator-style product advice

• One trained to give Alex Hormozi-style marketing advice

  ---

• Watch the whole episode:

• Listen on Spotify:

  ---

• Listen on Apple Podcasts:

  ---

Building Leverage Podcast: Sabrina Romanov

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

A serial entrepreneur who started with lead generation, is now a leader in AI product and building knowledge. He learned how to use AI as a tool for his success and his goal now? Share what actually works and help everyone have a better chance at achieving their goals with the help of AI.

The Guest

Cooper Simson — An entrepreneur and Ironman athlete who started a successful jewelry business with his now fiancé and sat at the forefront of Dan Martell’s Ventures.

Cooper started with a lead generation company. When things took a turn and lessons were learned, he and his now fiancé started their jewelry business Mogano. He then began working with AI and ended up becoming employee one at Martell Ventures helping take it from zero companies to fourteen at the time of his departure.

The Insights

Validate the idea before you build!

The AI era has flipped the script. Before, building was the bottleneck so our ideas were funded by faith. But now, thanks to AI, building most of the time is a trivial detail meaning the only real signal for success is if someone will pay you for your idea. Cooper’s hard rule working with Martell Ventures was pre-sell your idea to ten customers before you write a line of code. Validate the idea then build the success.

Speed is the #1 trait setting the winners apart from everyone else

Joshua’s best question was, "Out of the 3,000 startups you were able to overlook... what's the number one trait of founders who win?"

Cooper’s qualified response? "…speed and selling before you're ready." Out of the three thousand startups, it’s the ones who shipped before they were ready and iterated in public that proved themselves the winners. AI has truly compressed the build cycles to days and reducing the gap between thinking up an idea and actually shipping it is how you succeed.

The zero-to-a-million playbook

What’s Cooper’s zero-to-a-million playbook? Forget the mass-market SaaS dreams. Go find one person with a specific pain they'll pay real money to solve, then go find 50–100 more like them. Having AI like Claude doing the heavy lifting means you can researcher, build, and begin solving this pain in as little as a weekend. Then BAM you’re already making money in the thousands before you’ve had to hire anyone!

Cooper mentions the meeting recorder… there is already so many out there. This isn’t a specific pain anymore. (like he said though, no offence to anyone out there making a meeting recorder)

The core skill is no longer execution, its orchestration

Since our grandparents have been in the workforce we’ve been trained to execute tasks, but AI can now execute task sometimes better and definitely cheaper. Knowing what to ask for, in what order, and with what context are now basic skills everyone should have instead of just the executives they were formally reserved for. The people that can learn to orchestrate AI well will have the advantage; it’s the people who don’t that will inevitably fall behind faster.

The Window of opportunity is now and its closing

Every weekly Anthropic or OpenAI update makes those early users meaningly more effective.

Cooper's own prediction is a 4–5 year window, where being early is a massive advantage. The "calculator moment" analogy applies to this very well! The people who refused to use calculators got left behind, and the people who used them without understanding math got left behind too. The winners learn the tool, the fundamentals, and how to use them in tandem. The same scenario is repeating with AI.

What Is Cooper’s Top Claude Skill?

Cooper’s clear favorite was definitely Claude’s Last 30 Days skill. He mentioned it's hardcoded into pretty much every agent he runs — anytime one of his workflows needs to do research, it routes through that skill by default.

Words of Wisdom

“Just say yes to stuff. Opportunity comes all the time but most people either don’t see it, or they don’t say yes to it.” — Cooper Simson

Opportunities come up all the time but a lot of us forget they also go, and they have no problem leaving without us. It’s important to grab those opportunities that can better us before they go and if they don’t work out, oh well! Like Cooper also says, “The only thing that I keep doing is every time you get kicked in the balls, just get back up and try the next thing.”

The trap everyone tends to fall into is, they’ll read or watch something like this, feel all motivated, and then start planning. But the shipping never comes. So open Claude Code, build the smallest possible version of something today, and then start shipping next week. Grab your opportunity and say yes.

Your Next Move

Go search for a list of specific problems that someone will pay you to solve for them. Install Claude Code and a good voice-to-text tool. Pick a lane, content or product. Cooper did content to promote his AI work — he picked one primary lane. If your lane is content, save 30 posts within your niche this week and start posting next week.

Cooper did just this. Whenever he scrolls Instagram and sees content he likes in his niche, he saves it to a folder. An agent scrapes that folder, transcribes, identifies hooks, pulls metrics, and tags how and why that video preformed for its creator. Then when he’s ready to make a video, he gives Claude a topic and it goes through the database. He’s grew to 37,500 followers and built an email of 10k people with AI writing it all.

If your lane is product, talk to three people in your target group, find a $500/month pain, get a single person to pay you, then do the work manually with Claude.

Mentioned In This Episode

• GPT 2.0 — The early developer-tool version he used for his grant-writing business

• Stitch — The design tool that forced him to redo his whole design workflow

• Claude Skills — Last 30 Days, Last 24 Hours, Banana Claude, Humanizer, Notion Setup, his personal WAP Course skill

• Claude's Chrome extension — Cooper’s hack? Let Claude execute browser tasks instead of just describing them

• Nano Banana — For image generation, used via the Banana Claude skill (honorable mention)

• Dispatch — Anthropic's new phone-connection product

• Notion — For Cooper’s content database, managed via the Notion Setup skill

• GitHub — Fitness app… just kidding, Cooper uses for skill installation and private repos for his community

• Telegram — Where his jewelry inventory bot currently lives

• Grok API — Required for the Last 30 Days skill to pull Twitter/X data

Watch the whole episode:

Listen on Spotify:

Listen on Apple Podcasts:

Apple Podcast Building Leverage #10 - Cooper Simson

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.

Building Leverage #9 — Maverick Maltin

A 19-year-old kid skipped his summer internship, picked up his iPhone, and posted a video about AI that got 200 views. Eight months later he’s partnered with OpenAI, gains 150,000 followers from a single post, and turns down five-figure brand deals because the product doesn’t fit his audience.

The Guest

Maverick Maltin — 19-year-old AI creator with 800K+ followers who makes beginner-friendly AI content from his iPhone.

Maverick grew up traveling 50 countries before age 17. He started posting AI tutorials while his college friends were doing summer internships. Now he pulls 50 million views per month and works with OpenAI, Starbucks, and Canva.

The Insights

He found the gap everyone else walked past

Maverick noticed that AI content was split into two camps: technical breakdowns for developers and hype posts with no substance. Nobody was making content for the person who just wanted to learn the basics. He studied creators who were blowing up, spotted the hole, and filled it with simple tutorials filmed on his phone.

150,000 followers from a 20-second video about “secret codes”

His most viral recent video was a list of shortcodes you can type into ChatGPT. He filmed it in 10 minutes. The hook was seven words: “Here are five secret codes for ChatGPT.” No fancy editing. He added screen recordings and some typing sound effects. The insight: small, implementable tips outperform complex tutorials every time.

He never negotiates his rate

Maverick’s brand deal philosophy is blunt. His rate reflects the time, effort, and audience he built. He packages discounted long-term deals, but one-off rates are fixed. He’s turned down substantial offers from companies that disrespected his work in the negotiation. If a brand’s first email is calculating his CPM and telling him his views are low, he walks.

Reposting old videos is free growth

He reposts his own content every 60-90 days. His logic: if you gained 100K followers since the original post, 90% of your audience never saw it. He reposted a video from November and it performed. Instagram doesn’t punish it if you wait long enough. Most creators are too scared to try this.

He gave up selling digital products because his audience called him out

Maverick was selling guides and prompt packs. Then DMs started coming in: “You said this was free in your video. It’s $50.” That stung. He pulled everything behind a paywall and made it all free. Now he lets brand partnerships fund his education mission. His line: “All these people selling courses for $97 and $4.99, but it’s all free now.”

Best Quote

“You’ll never get shit on by somebody who’s doing better than you.”

Your Move

Open your Instagram, go to your liked/saved posts, and find 5 videos in your niche that got over 500K views. Write down the hook for each one. Film your version of the best one this week. Change the hook, add your personality, and post it. Maverick built 700K followers doing this. His buddy in Australia copied the same playbook and hit 10,000 followers in 30 days.

Mentioned in This Episode

• ChatGPT — Maverick’s primary tool for brainstorming and creative tasks

• Claude / Claude Code — His pick for writing, emails, and document design. Called it the most underrated AI tool.

• Poppy AI — His content workflow engine. 200+ video scripts, runs on Claude 4.6, replicates his voice

• Lovable — Top 5 AI tool

• Nano Banana (Google) — Preferred image generation model

• Perplexity — Go-to for research

• Gamma — Best AI tool for presentations

• Higgsfield AI — The tool behind faceless AI avatar accounts (the $10K/month monk blueprint)

• ManyChat — Key to Instagram growth through comment CTAs and freebie delivery

• Manus (Meta) — His prediction for the next big AI platform

Watch the full episode:

Listen on Spotify:

Listen on Apple Podcasts:

You’re getting this because you care about building with AI, not just reading about it. If this landed, forward it to one person who needs to hear it.