The honest reason we do this: we’re inundated with hot takes, thought leadership, and endless feeds. While there’s no shortage of content (especially with AI), most of it is noise. This is the short list of what we actually found useful - the links we send each other and forward to our founders. Each partner picks a single thing from their week and writes a couple of paragraphs on what it is and why it’s worth your time.

We’re still figuring out what this becomes, so reply and tell us what’s working. And if it’s useful, feel free to forward it to other founders.

And with that out of the way, let’s get into it...

Mark — AI Models Are Getting Abliterated

https://flare.io/learn/resources/blog/ai-models-cybercrime-quantitative-analysis

Railway’s breach report said the attacker had “access to abliterated models and substantial compute resources” - but you may be wondering what an “abliterated” model is. Abliteration is the cheap version of removing a model’s refusals: find the single direction in the weights that carries “no,” project it out, done. No retraining, no fine-tuning dataset, minutes on one GPU, and most of the model’s capability survives. It also requires open weights, which is why a closed model behind an API can’t be treated this way.

Flare went looking for the demand side and found the conversation up 745% from January to October, from roughly 11 to 93 unique posts a month across 100,000+ Telegram channels and 100 dark web forums. The market is still mostly theater, most of what’s actually for sale is resold or stolen API keys rather than weights. GLM is the exception and the only model selling above list price, at 2.6x, because standing up an abliterated GLM is genuinely hard. The damage is real though: ARTEX running on DeepSeek and GLM-5.3 got into at least nine South Korean banks, and Shinhan exposed data on about 25,000 customers. If refusal is one vector in the weights, then the guardrails on a downloaded model are an speed bump, not a wall.

Moataz - Why Your Smart TV Is So Cheap

If you’ve wondered how a 65-inch TV costs less than a nice dinner, this thread explains it. A researcher who rooted his Samsung reported an always-on mic, geolocation, network mapping, and automatic content recognition, which screenshots whatever is on screen every 500ms, including your laptop or console over HDMI, and sends it home for ad targeting. The economics explain why: in Q3 2024, Vizio lost $6.7M selling TVs and made $115.8M in gross profit from ads and viewing data. The hardware is the customer acquisition cost.

The thread also has a practical 15-minute fix: turn off ACR, put smart devices on their own network, and filter trackers at the DNS level. That second step matters more every month. These devices rarely get patched, and as AI makes finding exploits cheaper, an unpatched TV on your home network is an open door.

Farzad - Clinical AI: Who captures the value?

https://bipartisanpolicy.org/article/paying-for-clinical-ai-principles-for-policymakers/

Clinical AI has the potential to fundamentally change the economics of healthcare by dramatically reducing the marginal cost of delivering care. But our reimbursement system was designed around human labor, not software that can autonomously diagnose, treat, and manage patients. Under fee-for-service, AI could actually increase healthcare spending by driving more utilization, while value-based care creates an opportunity to align AI’s efficiency with better outcomes and lower costs. For founders building clinical AI companies, reimbursement and business model design need to be foundational decisions, not afterthoughts. Who pays, who captures the savings, and who assumes clinical and financial risk will ultimately determine which companies build enduring businesses. The winners may not be those with the best clinical models alone, but those that figure out how to translate clinical intelligence into economic value.

Ramy - Aligning Startups With Their Chronically Anxious AI Investor
https://www.linkedin.com/pulse/aligning-startups-vcs-ramy-adeeb-lhzwe/

In 2008 I sat in Andy Rachleff’s GSB class and wrote down the most important lesson about startups: customer pull beats everything.

But today most founders are finding themselves building for a second customer, the chronically anxious AI investor. Worried about missing out on the next Anthropic and becoming irrelevant, they’re applying the same growth curve to every market. Not surprinsgly, founders start optimizing for what gets funded instead of who is actually pulling.

But the best founders I know are still quietly following the first customer (while tailoring the story to the second). It takes a bit of effort, conviction and convincing. 18 years on, Andy’s class still feel more useful than most of what’s circulating now.

The post-Covid boom had a problem: too much capital funded too many companies that shouldn’t have raised. Today we have the opposite problem. The flood of capital into AI is leaving many companies that should get funded without it.

We see it in our own portfolio. In the past year, every company that has raised a subsequent round did so less than three months after its pre-seed, and at eye-popping valuations. Nearly all were tied to the AI infrastructure trade or sell tokens or services. There is increasingly nothing in between.

Three things are driving this.

1. Venture has become a game of capital absorption

There’s an old joke in venture: come for the carry, stay for the fees. And the fees have never been better! But when you manage tens of billions, the question shifts from what multiple a company can return to how much capital it can absorb. Data centers, chips, inference sold below cost, data suppliers to the frontier labs. Each can consume billions, which is exactly what a giant fund needs.

That logic trickles down the venture ecosystem. Series B investors chase the companies they think can raise mega-rounds quickly. Series A investors chase the companies they think Series B investors will chase. You’re an obvious winner almost immediately, or you’re invisible. This is all exacerbated by GPs fear of becoming irrelevant if they’re not in the right names or if the portfolio has not exited when the music stops.

2. AI has broken the growth benchmarks

The concentration at the top creates another problem further down the stack: AI has completely reset investors expectations around fast growth. The infrastructure buildout has produced companies going from zero to hundreds of millions in revenue faster than anything we’ve seen since the telecom buildout of the late 1990s.

Not all revenue is equal, though. Some of these businesses sell to the same five frontier labs. Many have negative margins once inference is counted. So many are services businesses valued like software, or count pilot budgets as recurring revenue. Many would not survive public-market scrutiny, the same way e-commerce brands valued at 10x revenue in venture were repriced near 1x once they went public. Still their growth becomes the benchmark for everyone else.

As a result, an AI-native software company going $0 → $2M → $10MM, with passionate customers and solid margins, gets tossed around. By any historical software standard that’s a top-decile trajectory for a Series A. VCs would have killed for it three years ago.

3. The winner-take-all fallacy

But there is a deeper assumption underneath all of this: that AI will be winner-take-all. If a handful of companies capture most of the value, overpaying for the likely winners is a smaller mistake than missing them.

We’ve heard this before—with railroads, telecom, the internet, mobile and cloud. But AI comes with more creative ideas for why this time it’s different. First it was AGI, yet whether or not AGI has arrived, model competition is fiercer than ever. So today’s term is RSI, recursive self-improvement: once a model can improve itself, the leader’s advantage “compounds” (a favorite word in the AI lexicon) until everyone else is hopelessly behind.

But recursive doesn’t mean exponential. You can recursively improve your way toward an asymptote, and that’s a fair description of what we see today: more compute and capital producing better models with diminishing gains, while competitors are catching up at an ever faster pace. For RSI to crown a permanent winner, each generation would have to get better at improving the next one while making it harder for rivals to follow. There’s little evidence of either.

To be clear, there are real winner-take-most businesses: Network effects, distribution, economies of scale, and market places do create winner-take-most businesses. But AI is a technology, and technologies don’t have market structures. Businesses do. And it’s not driven by capital spend. The railroad builders spent fortunes laying track, and many of them lost money. The businesses that used the track did quite well.

This is the part I think we’re missing in AI. Hundreds of billions of dollars are being spent to make intelligence better, cheaper and more abundant. If that works, the cost of software’s most important new input keeps falling, and that should create more companies, not fewer. Venture may be financing the commoditization of the very thing it treats as scarce.

So what do you do about it

It depends on where you’re sitting.

If you’re building AI infrastructure, take the money.

This is one of the greatest capex cycles we’ve ever seen. Enjoy it but don’t confuse a great market with a permanent one. Eventually capacity catches up and prices fall. Cisco is probably the most useful cautionary tale. During the internet buildout, Cisco briefly became the most valuable company in the world. The internet thesis was completely right. The assumption that those infrastructure economics would persist was not.

So raise while capital is available, take secondaries (a lot of secondaries) when you can, and entertain any offer that prices today’s economics as permanent. Learn from Cursor who saw their market share dropping rapidly but were quick to land a spectacular deal.

If you’re building applications, stay alive.

If you have diversified customers, good retention, solid margins and 3x yoy growth, nothing is wrong with your business. The market today is pricing a different game, but venture capital appetite turns on a whim. Keep growth healthy and burn under control. Make sure you own the system of record or a critical workflow for your customers. Don’t contort the company into whatever raised $200 million this month, whether that’s selling data to the labs or deploying an army of forward-deployed engineers to generate services revenue that hides a lack of product-market fit. Treat margin as an asset, because eventually investors will care again.

Your defense against the labs is what they don’t have: deep workflow integration, proprietary customer data, and distribution into buyers who don’t take calls from model companies. And when the over-building is done (we always overbuild), your most important input will get cheaper.

At the height of the dot-com and telecom boom, Netflix offered to sell itself to Blockbuster for $50 million and was turned down. A few years later, it built a streaming business on the cheap bandwidth the boom left behind, and today it’s worth $300 billion.

You don’t need to win the infrastructure race. You need to survive it.

If you’re allocating capital, don’t make the same bet twice.

This may be the biggest implication. Most institutional portfolios already have enormous exposure to the AI infrastructure trade: Nvidia and the hyperscalers in public equities, frontier labs through growth managers, and the same consensus names through venture managers. A new fund chasing those companies is the same bet expressed through a different vehicle, now with mounting debt and stretched valuations attached.

There is another way to underwrite the AI opportunity. You can believe AI becomes one of the most important technologies of our lifetime without knowing which model wins. In fact, if intelligence becomes abundant and cheap, the number of businesses that can profit from it should explode. You can also be extremely bullish on AI and simultaneously believe that the AI buildout is mis-pricing many of the companies best positioned to take advantage of AI.

The infrastructure bet requires you to be right about who wins. The application bet requires you to be right that AI works. Those are very different bets.

The paradox

That’s what makes this moment so strange.

We are spending unprecedented sums to make intelligence cheap and abundant while underfunding the companies best positioned to use it. When an important input gets radically cheaper, history says the biggest returns go to whoever figures out what to do with it. For application (and dare I say SaaS) startups, this isn’t the end. It’s a whole new beginning.

There’s a story Silicon Valley really wants you to believe right now.

It goes like this: intelligence itself is the scarce resource. Whoever builds the most powerful model will, in Sam Altman’s words, “capture the light cone of all future value in the universe.” The argument goes even further: that the first company to reach AGI will permanently leapfrog everyone else. Everything else — the apps, the workflows, the industries — becomes downstream. Just features waiting to be absorbed.

It’s an elegant story that’s supported by the shortage in inference capacity. It is also a very convenient one for the hundreds of billions of dollars now concentrated around a handful of companies.

But it’s mostly hubris. And we’re betting against it.

Models are commoditizing fast

Two years ago, GPT-4 class intelligence cost roughly $30 per million tokens. Today, similar performance is available for under a dollar. Meanwhile DeepSeek, Kimi, and Qwen keep closing the gap at an even smaller fraction of the cost.

None of this should be surprising.

Compute commoditized. Bandwidth commoditized. Storage commoditized. None of those looked like commodities when they first appeared either. They looked magical, scarce, and impossible to catch. Then markets did what markets always do: competition increased, supply expanded, and differentiation collapsed. Intelligence is heading down the same path.

Infrastructure captures pennies on the dollar

The historical parallels are everywhere. The railroads didn’t dominate the industrial economy. Standard Oil did. AWS created enormous value, but the defining companies of the cloud era became Stripe, Shopify, and Snowflake.

The AI-native companies that define the next decade probably won’t win because they have a marginally better model. In fact the models are already good enough to transform most industries. They’ll win because they own the customer relationship, the proprietary data, and the workflow that’s too painful to replace.

At Tellme Networks, one of the first voice enterprise AI companies, we were making over $120M in revenue automating call center for enterprises. We didn’t have our own voice recognition engine. Instead we used Nuance, the most advanced provider at the time, and paid them pennies on the dollar. They were competing with us on every contract—and losing. Because we had the best applications targeted for each vertical.

Short on hype, not ambition

More importantly, the most important AI companies probably haven’t been founded yet.

Google wasn’t obvious in 1993. Amazon wasn’t obvious in 1996. The founders building the defining companies of the AI era are likely out there right now, working on problems that still sound too niche, too messy, or too early. We don’t know who they are yet — and neither does anyone else

But we’ve also seen enough cycles to know that the most important companies are never built by the people who own the infrastructure. They’re built by the people who can’t stop thinking about a specific problem nobody else thinks is worth solving yet.

That’s the long AI bet.

A brief history of hype

Let's examine some examples from the past 20 years.

• When Google was raising its seed round in 1999, the prevailing wisdom was that "search is dead" since none of the players could monetize. So uncertain were Google's founders about monetization that they offered to sell the company to Excite@Home for $750K--but the latter backed away. (The hottest technology meanwhile was bringing the internet to the phone using voice recognition with Tellme, my alma mater, raising a whopping $250M in VC capital. The company exited to Microsoft years later for $800M, a tiny fraction to Google’s IPO)

• When Facebook was founded in 2004, the hottest trends were smart fridges and RFID, all of which went haywire. Meanwhile investors were running away from social networks following the blunders of Friendster.

• When Uber and Airbnb raised their seed rounds in 2009, the hottest trends were Social (buoyed by Facebook’s success) and cleantech, a sector that arguably caused the demise of Kleiner Perkins v1. No one believed the "sharing economy" would actually work. In fact, that term didn't gain mainstream adoption until three years later. Investors then dumped money into other sharing economy opportunities, most notably electric scooters, where everyone lost their shirt.

• Perhaps no better case illustrates his trend than the most recent mega-exit, Figma, which was acquired by Adobe for $20B last year. The collaborative design startup raised its seed round in 2013 to leverage websockets, the technology enabling continuous communication between the browser and server.  But, websockets was far from being the hottest technological topic of 2013. That award went to Google Glass. So exciting was the new technology that Andreessen Horowitz and Kleiner Perkins launched a dedicated Google Glass fund in April 2013, certain that “developers will create thousands of ways for millions of people to use Glass and improve their lives and the world.” 

The story continues–just with different technologies reaching peak hype from Internet of things to the Sharing Economy to 3D Printing to Electric Scooters to Autonomous Vehicles to AR/VR to Crypto and now, dare I say, to LLMs! 

The merits of hype

Hype is good for innovation: it attracts capital, draws talent, and raises awareness from prospective customers.  But inevitably expectations get out of touch with reality and the hype bursts. What happens next depends on user adoption or, put differently, the technology finding product market fit.

When it comes to what became "novelty technologies" such as Google Glass, 3D printing, and scooters, the bust signals the end of the story. These technologies fit well into the writer John Thackara's observation: “If you put smart technology into a pointless product, the result will be a stupid product.”  But when the technology has meaningful applications, a bust is only temporary. Examples include the first internet bubble in 2000 and autonomous driving in 2019/2020, both poster children of Amara's law: “We tend to overestimate the effect of a new technology in the short run and underestimate its effect in the long run”

But predicting which way things will go is extremely challenging. Technology alone does not create the future. Instead, the future is the result of an unpredictable mix of technology, business, product design, and culture. Facebook had a poor choice of technology (PHP), yet it succeeded thanks to a brilliant product. Google took off because of its search technology, but its success came from a clever business model innovation with inline ads. All of this seems clear in hindsight, but it is entirely opaque looking forward.

Hype and returns 

Although hype inevitably bursts, chasing hype can be a fine strategy for late stage funds. It enables investors to tell exciting stories of “platform shifts” and “once in a generation opportunities.” and feels safe with infinite promise and abundant capital. And it can generate handsome returns if a manager can successfully pick the winner (or, for investors with enough clout, if they can hype their own company as the winner and orchestrate a timely exit)

For early stage investors, however, investing in hype is extremely challenging. The reason: by the time a sector is hot it’s too late for a new entrant to be successful. As a result seed funds do not get the opportunity to pick a winner in an already hyped sector. Perhaps the best analysis here comes from Andy Rachleff, a co-founder of Benchmark and a mentor to many of us in the venture industry. Analyzing years of early stage venture performance, and adopting a framework from his own idol, Howard Marks, Rachleff deduced that the only way to generate outstanding returns at the early stage is to be right and non-consensus. 

Andy’s insight also presents a challenge for investors: you only know if you're non-consensus when you make the investment, and you don't know if you're right. However, what signal can investors look for? Or, more specifically, what did Google, Facebook, Uber, Airbnb and Figma all have in common? They certainly leveraged an inflection point in technology–albeit one that wasn’t necessarily obvious–like PageRank, GPS in cell phones, and Websockets. They had excellent teams. But, most importantly, they solved a need that people turned out to be desperate for. 

For early stage investors, success isn't about predicting the future of how technology evolves. Rather it’s about finding signs of a clear market need (that may appear nascent or even solved by existing players) and backing teams who can capitalize on these needs. In short, when it comes to returns, product-market-fit eats hype for lunch. 

(Special thanks to Andy Rachleff and Samit Kalra for their invaluable feedback in writing this post)

Much has been said about SVB’s failure. As a client of the bank for over two decades, first as an entrepreneur and then an investor, the impact of its failure is comparable to what Michael Moritz so eloquently described as “a death in the family”: something to mourn and reflect on. By now the structural reasons behind its failure are well known. What interests me is the role it played in the technology ecosystem, and the role technology played in its demise.

The bank for founders

Startups are a hard bunch to bank: there are no assets to lend against. And many fail, often in spectacular fashion. SVB understood these needs. The bank underwrote against the risk profile of startups better than most, and leveraged its partnerships with venture firms to become the partner of choice for Venture Debt. Not surprisingly, startups and funds across the globe flocked to bank with SVB.

Fortunately, for the tech ecosystem, other institutions took notice. JPM recruited many of the failed bank’s leaders in recent years. And the Venture Debt asset class is now offered by many companies such as TriplePoint, Comerica, and Bridge Bank. Newer startups meanwhile have been flocking to Brex and Mercury, attracted by their modern products and founder-cenric perks. SVB was the first bank built for founders, but by the time of its collapse this month it was far from being the only one.

Herd mentality ..

Still, SVB’s failure was epic. During the last major bank run on Washington Mutual in 2008, $15B of deposits were withdrawn in ten days. In the case of SVB, $42B of withdrawals were attempted in just a few hours. This frenzied behavior merits a pause. Herd mentality is an ancient human pulse. But technology put this impulse on warp speed.

When depositors mainly consist of startups backed by the same group of venture capitalists on their boards, and these venture capitalists all call their companies to let them know that others are withdrawing their funds from SVB, you end up with a textbook definition of a bank run. This was foolish. As Hemant Taneja of General Catalyst wrote: “causing a bank run at the most important institution in your sector can hardly be considered an act of fiduciary duty”

.. enabled by Technology

I am also surprised by how little attention has been given to the role technology played. This was truly the first Twitter-fueled bank run aided by mobile banking. Admittedly, SVB had the most tech-savvy clients today. But in a few years every bank will have depositors who are social media-obsessed and have two-factor authentication. The next bank failure will likely have a root cause other than government bonds held to maturity in a rising rate environment. But how it fails will likely look similar.

The need for circuit breakers

The FDIC were quick to respond by guaranteeing deposits and announcing a program shielding other banks from the risks of rising interest rates, the problem they understand and control. But they also need to re-instiute the friction in bank runs that technology has all but removed, for example via circuit breakers similar to the ones in the stock market.

The road ahead

The downfall of SVB will likely result in further tightening of the venture debt industry, making an already difficult year for funding even more challenging. But this will also create an environment where only the strongest and most innovative companies survive, leading to stronger startups with more profitable business models. SVB’s demise was a shock to the system, but sometimes a shock can be a good thing.

Even though toilet paper is out of stock, manufacturers are not increasing supply. Having been in the business for a long time, they know consumers hoard toilet paper in times of crisis — it’s happened before during Hurricane Katrina and the fallout from 9/11. They also know the scarcity is followed by an immediate decline in demand as consumers go through the extra rolls filling up their cabinet. Seeing the shortage, a reactive investor might rush to invest in paper mills. A more savvy investor knows not to. It boils down to a simple fact: The rate with which we go to the toilet has not fundamentally changed.

In times of crisis, basic business indicators can become extraordinary — almost unrecognizable. Grocery stores are reporting 80% week over week growth. Travel, hospitality, and real estate transactions have come to a standstill. Should venture capitalists back online grocery experiencing record growth or a travel startup that’s unlikely to have any meaningful revenue for the next quarter? The answer is nuanced.

By plotting the long-term impact of the crisis against short-term impact, an interesting picture emerges. On the bottom left of this grid are industries suffering both in the short and the long term. Think Macy’s, already in decline and now forced to shut its doors for the foreseeable future. The crisis has likely accelerated an inevitable demise.

The top right quadrant is equally obvious. Zoom use is at an all-time high. E-commerce, video conferencing, and telemedicine all fall in this bucket: trends accelerated by the crisis. It’s a good place to already be an investor. But new investors must contend with frothy valuations.

The remaining two buckets are more nuanced. At the bottom right are industries that will benefit in the short term but will not be affected over the long term. Our toilet paper story is a prime example. Food supply, ventilators, and many health care products fall in that bucket. There may still be good opportunities for a public investor, but less so for venture investors. A few startups in these industries may see a short-term cash infusion; most won’t even launch till the glut is over.

The last quadrant, on the top left, is arguably the most interesting. The companies in this quadrant will suffer in the short term but will outperform the competition over the long term. A good example is Airbnb. While the business is facing a major drop in bookings, Airbnb’s asset-light model and the likely demise of many traditional hotels points to future dominance. Not surprisingly, while the CEO of Marriott issued a heart-wrenching plea to his employees, savvy venture investors went to Airbnb asking if it would raise capital.

If history is any guide, it’s the upper left quadrant that will likely see disproportionate venture returns. Google raised its financing just as the first tech bubble burst and investors were avoiding search companies like the plague. And Uber and Airbnb both were founded during the 2008 financial crisis when global travel was down 25%. At a time when cash is not readily available to most companies, investors may be tempted to only support these companies flush with revenue. But those with the stomach to focus on the long term will reap disproportionate rewards.

When YC first introduced the SAFE agreement in 2013, few expected it to become the de facto instrument for early stage investing. But the document has become all but ubiquitous, with some unintended consequences. To address the expansion and the shortcomings, YC just unleashed a revised document dubbed post-money SAFE. The changes are subtle — but have significant repercussions to founders and investors alike. Below are the major changes and my take on each.

1. The new SAFE valuation is post-money

What it means

The valuation cap in the new SAFE is post-money (as opposed to pre-money). For a company raising just one SAFE round, there’s effectively no repercussions: an investor willing to invest $2M on $8M pre-money is presumably willing to invest $2M on $10M post-money, with the same resulting ownership of 20%.

But things are different when multiple SAFE rounds are involved. Imagine the company raising an additional $2M on the same note a year later. Under the old SAFE agreement, the company will have effectively raised $4M on $8M pre / $12M post with investors owning 33.33% and founders 66.66%. Under the new agreement, the company will have effectively raised $4M on $10M post with investors owning 40% and founders 60%. Why the difference? Because in post-money agreement, subsequent SAFE investors dilute the founders but do not dilute the existing SAFE holders.

The mechanics here are the same as with priced round: if an investor and company agree on a pre-money valuation, then an investor’s resulting ownership is a function of their investment, the pre-money valuation and the overall round size. But if an investor and company agree on a post-money valuation, then an investor’s resulting ownership is a function of their investment and the post-money valuation irrespective of the round size. SAFEs have just become the latter.

My take

As YC’s blog post pointed out, the post-money SAFE has a significant advantage for both founders and investors — the ability to calculate precisely how much ownership of the company has been sold. But the new mechanism is a also boon to the early SAFE investors who are no longer getting diluted by additional SAFE rounds. How this change affects valuations for subsequent SAFE rounds remains to be seen¹

2. Investors no longer get pro-rata rights by default

What it means

The old SAFE granted SAFE holders a pro-rata right to protect their ownership. The new SAFE does not. Instead, company and investor can enter into a side letter granting the pro-rata right to investor. To make things simple, YC published a standard side letter on its website.

My take

Companies and investors will now negotiate pro-rata rights on a deal-by-deal basis. Investors writing a larger check will have more sway. In reality, this is already the case: even though previous SAFE holders were granted a pro-rata right, the right was often revoked in the Series A Investor Rights Agreement for investors not meeting a “major investor” criteria. The new mechanism is more in line with market realities.

3. The new pro-rata right takes place in the Series A, not the Series B

What it means

This is an important technicality that merits a deep dive. Let’s assume the SAFE converts in the Series A. Under the old SAFE, the SAFE holders — now holders of Series A Preferred Stock — were diluted by other SAFE holders and by the Series A, but were then allowed to purchase a pro rata portion of the Series B round to maintain their ownership. Under the new SAFE, if the investor signs the pro rata side letter, the investor is granted the right to maintain their ownership in the Series A round. In other words, an investor putting $2M on $10M post SAFE can effectively retain that 20% ownership through the Series A and possibly beyond.

My take

The old scheme granted investor pro rata rights one round removed. It didn’t make lot of sense. Not surprisingly, many investors would amend the SAFE agremeent with an explicit Series A pro-rata clause with a pre-defined ownership percentage. The new scheme is a more logical and transparent approach.

4. SAFE holders now more closely resemble equity holders

What it means

SAFE holders are now explicitly granted a number of rights similar to equity holders includeing the right to receive dividends (ICO proceeds anyone?) and the right to receive the same compensation choices in a liquidity event as equity holders. Additionally, the new SAFE introduces a new section, 1(d), that spells out the liquidation preferences following an acquisition in detail with SAFEs on par with preferred stock.

My take

This is sensible. As YC pointed out: SAFEs are “really better considered as wholly separate financings, rather than ‘bridges’ into later priced rounds.”

5. The new agreement is supposedly not modifiable

What it means

The new SAFE adds the following line: “This Safe is one of the forms available at http://ycombinator.com/documents and the Company and the Investor agree that neither one has modified the form, except to fill in blanks and bracketed terms.“

My take

Hallelujah! The S in SAFE stands for Simple. But attorneys constantly make edits to the document to justify their value add (and boost billable hours when none were needed). It’s one of my biggest pet peeves. And the resulting back-and-forth defeated the entire purpose of the SAFE.

The new SAFE attempts to address this issue by implying the document should not be modified — presumably without removing this paragraph first. Let’s hope it works. Somehow I suspect that company and investor counsel will simply drop this line and edit the SAFE as they please, or contribute their nuggets of wisdom in the pro-rata side letters instead. (A special place in hell will be reserved for those who edit the document without removing this line!)

6. Investor payout in a liquidity events is streamlined

What it means

Under the old agreement, if an acquisition took place before the SAFE converts, investors had the right to choose between receiving a cash payment equal to their investment amount or convert to common stock. Under the new agreement, investor automatically get the larger of their investment amount or the proceeds from the acquisition as if they had converted to common.

My take

Acquisitions taking place before a SAFE converts are often talent acquisitions — where the acquirer cares less about paying out investors and more about retaining management. Often it is the entrepreneur fighting with the acquirer to ensure investors are paid out. SAFE holders, not being shareholders, have little say in this transaction. The new mechanism is simpler and more transparent. But don’t expect it to alter this complex dynamic much.

7. A majority of SAFE holders can now amend the SAFE agreement for everyone

What It means

The old SAFE agreement could only be amended upon the written consent of investor. The new SAFE allows amendments by written consent from a majority of SAFE holders. There are some boundaries: the company must solicit the consent of every investor (even if consent is not granted), the Purchase Price may not be amended, and all investors must be treated in the same manner in the amendment.

My take

There are legitimate reasons to amend the SAFE: just last month I experienced a case where the founder incorrectly filled out the discount rate as 10% as opposed to 90%. In another instance, a founder merged two SAFE documents and in the process dropped the cap clause making it effectively an uncapped note. To amend these SAFE, founders had to get everyone’s consent which was no simple task. This clause makes it easier for founders to amend the SAFEs.

But the new clause also allows a majority of SAFE holders to enforce substantial changes on all the holders — including the actual cap. When companies struggle, it could be used by larger investor to drag-along smaller investors into accepting punitive terms.

8. Non-accredited investors are really not welcome

What it means

The old SAFE stipulated that the investor must be an accredited investor under Rule 501 of Regulation D of the securities act (which defines an investor, among other things, as having $1M in assets or $200K in income) But it never spelled out the repercussions. The new agreement explicitly grants the company the right to void the SAFE and return the Purchase Amount if the investor is not accredited.

My take

SAFEs have effectively become equity. Requiring participants to be accredited investors makes sense. And for Bay Area residents, $200k is barely enough to pay rent let alone invest in startups.

Conclusion

The first version of a new invention often resemble their predecessors with some awkwardness— the first automobile looked like a horsecarriage with an engine. Similarly, the first SAFE resembled a convertible note. It had complex pre-money conversion math. It talked of termination and expiration — even though there was no mechanism for expiration. It was meant as a bridge instrument and it read like one.

The new SAFE has matured into its own isntrument. Its emphasis on transparency should be lauded. And, while not perfect, it’s a welcome upgrade on many fronts to a document that has quickly become the standard in the industry. YC seems to have followed its own advice: build something people want. Then make it a little better!

Footnote 1: Detailed calculation comparing pre-money and post-money SAFE conversion and their impact on valuation

Imagine a company that has issued two SAFEs

• A $2M SAFE with a cap of $8M pre (old SAFE)/ or $10M post (new SAFE),

• A year later, the comapny would like to raise an additional $3M SAFE. The company has made progress and would like to increase its cap to $12M pre (old SAFE)/ or $15M post (new SAFE).

Let’s assume the company has 100 shares in its cap table held by the founders.

Under the old, pre-money SAFE agreement, the first SAFE receives 20% of new capital, or 25 new shares [25 new shares out of 125 total shares is 20%]. The second SAFE receives 20% of new capital, or 25 new shares as well. But the investors will dilute each kther: they each end up with 25 out of 150 total shares or 16.67%. The founders will end up with the remaining 100 out of 150 shares or 66.67%

Under the new, post-money SAFE agreement, the company will issue 40% of total new equity, which is 66.66 new shares [66.66 / 166.66 = 40%]. Each investor will receive 33.33 new shares and ends up with 20%. The founders will end up with 100 out of 166.66 shares or 60%.

What if the founders want to maintain their ownership at the same level held in the pre-money world (66.67%)? The founders can only issue 50 new shares. Because under the post-money SAFE later investors don’t dilute early investors, the first investor is getting 20% of the company, or 30 shares. The second investor can only get 13.33% of new equity (or 20 shares). Founder can either raise less money at the same valuation ($2M on $15M post = 13.33%), or raise the $3M at a substantially higher valuation ($3M on $22.5M post = 13.33%). How founders and later SAFE investors work this out remains to be seen!

Footnote 2: you can view the actual redline comparing both documents here

I am in my yoga class. The instructor introduces himself as a cis male. “Please correct me if I refer to you with the wrong gender pronoun,” he says. There’s a hint of apology in his voice — as if being a straight born male is an inferior form of being. I wonder if he actually cares about offending our pronoun- sensibilities. I lose my train of thought as I struggle with a variation of downward dog.

I step outside the yoga classroom. It’s a foggy morning in San Francisco. The ocean mist fills my lung — a welcome change from the stuffy smell of yoga mats. I walk half a block. I am overrun by a strong stench. On the sidewalk are three tents filled with homeless. I wonder why our city has more homeless than any other.

My Uber driver is from Afghanistan. He’s chit chatting about his live-in girlfriend and his wife back home. Life has gotten so expensive, he says, he now pays $2k for a two bedroom in Sacramento. He comes every Thursday and stays the weekend with his cousin. There’s more work in San Francisco. He is hoping he can get a girlfriend in San Francisco so he doesn’t have to stay with his cousin. But it’s too expensive.

The city might pass new rules about Uber, he frets, that will make it harder for new drivers to register. I read in the paper that the city “needs to balance the needs of all citizens, not just those who use ride hailing.” I wonder what PR firm the cab-medallion lobby hired this time around.

I am taking the BART from the mission to downtown. There’s a man sleeping in the back row. I read that BART officials voted against new security measures, following two homicides, because cameras are a threat to privacy. I wonder how many BART officials own a cell phone and if they know that, at any point in time, Facebook, Google, every telecom carrier, device manufacturer, a few dozen apps and a half dozen governments know their exact whereabouts.

Today is the opening of the Transbay terminal, a $2b public transportation hub. The building is stunning. And the park on top is a welcome reprieve from the streets below. It’s almost perfect. There’s a sign that Caltrain might stop there in 2028. I wonder if it’s a typo or a joke.

I arrive home. There’s a broken window in the car in front of our apartment building. I am desensitized to car break-ins by now. I head upstairs. There’s a post on Medium about leaving San Francisco. I understand why people would choose to leave. But of the millions of Americans who move every year, only ones leaving San Francisco write a treatise about their move — as if a social media post can conceal the inconsequentiality of our choices.

The sun is setting outside as the fog rolls in. This city I have called home for two decades has gotten as expensive as Tokyo and as seedy as the Bronx. Its summers are cold, its politicians inept. And its left-leaning inhabitants possess a punitive orthodoxy that is more reminiscent of Mullahs than liberals. But I chose to stay.

I chose to stay because the entrepreneurial spirit in this city is unlike any other. Because the optimism of Haight Ashbury and the fight for human rights this city pioneered before the rest of the world is still here. Because real tech is born within thirty miles from here. Because I prefer a city obsessed with Medium over one obsessed with Instagram. For that, and more, I choose to stay — and to write about it.

There’s a scene in The Big Short (a must-see movie about the housing crisis) when Mark Baum and his staff realize that a local Miami mortgage broker has been conning strippers into taking out highly leveraged mortgages. We haven’t quite reached that level of shady with the Bitcoin bubble. But we may not be far. An online search for bitcoins now serves up plenty of ads to buy the digital currency using a credit card loan.

Cryptocurrency advocates want you to believe we are not in a bubble. While some of these people are well-intentioned tech idealists, others are international miners engaged in price-fixing. So before investing in the hype, it’s worth knowing which claims are half-truths at best. Here are the top five myths Bitcoin promoters want you to believe.

Myth #1: Bitcoin has limited supply

You often hear that Bitcoin supply is limited — just like gold or the works of Picasso. While that’s technically correct (only 21 million bitcoins will ever be created), that sense of urgency isn’t so urgent as you’re being led to believe. Here’s the thing: You don’t need to buy a full bitcoin. Instead, you can purchase as little as one Satoshi, which is equivalent to 1 / 100,000,000 (one hundred millionth) of a bitcoin. In other words, we have 2,100,000,000,000,000 Satoshis, or 2,100 trillion units to trade. Compare that with 5.8 billion ounces of gold in the world, or 1,885 paintings by Picasso and you realize this supply is not so scarce after all.

Additionally, the Bitcoin community can ‘split’ at any point to create a new currency that shares the transaction history with Bitcoin up to a certain time and date but duplicating all future supply. Bitcoins has already encountered three forks this year.

Myth #2: Blockchain is what makes bitcoins so valuable

Cryptocurrency advocates point to blockchain as a game-changing new technology that justifies the price of bitcoins. The argument is flawed on two fronts. First, blockchain isn’t the panacea we make it to be. Imagine if email didn’t require a central server run by Google or Yahoo. Instead, you could email anyone directly without going through any central servers, but with one caveat: you must download the entire history of every email ever sent (to you or to anyone else) on your computer first and they must as well. It would eventually work, but your computer would come to a full stop before you could send a single email. That’s how the blockchain works. It’s decentralized, but also incredibly inefficient, slow, and prone to fraud. Secondly (and more importantly), bitcoins are built on top of blockchains. But you can create blockchains that have nothing to do with bitcoins. The two are not linked in value.

Myth #3: Bitcoins are secure

Each cryptocurrency has a private key (A very long number. In the case of bitcoin it is a 256 bit or a roughly 165 digit long number) that you need in order to sell it. But financial security isn’t just about protection; it’s also about remedy. Because there’s no central authority, and because bitcoin transactions are, by design, irreversible, those keys are a hot target for thieves. Not surprisingly, bitcoins are being stolen at an unprecedented rate (accounts are hacked; phone numbers are ported to new carriers then used to reset passwords, and an entire exchange was once taken down).

Many companies now offer to store your keys safely offline. Xapo, for example, keeps keys safe in secure bunkers in the Swiss Alps. It’s an ironic twist that the only way to keep digital currency safe is to store it in a bunker.

Myth #4: Cryptocurrency is mined in China because electricity there is cheap

This one is only half a lie. The cost of electricity in China is about 6 cents per kilowatt hour (kwh), compared with about 12 cents per kwh in the US. But electricity prices in the U.S. aren’t flat: they range from 45 cents in Hawaii to 7 cents in Washington state, almost the same price as in China. So why then is the US not a major producer of bitcoins? Because electricity is only part of the answer.

China mines over 60% of all new bitcoins today, followed by Georgia and closely thereafter by Venezuela. What do all these countries have in common? Capital controls. Bitcoin miners in these countries are willing to spend more money to move their money out of the country. What we have created with bitcoins isn’t democratized currency, but rather the world’s best tool for evading capital controls.

Myth #5: Bitcoin has intrinsic value

Bitcoin promoters create an analogy between bitcoins and historic stores of value like gold in the cost of production: it takes electricity to “mine” bitcoins. But price is not a function of production cost; it’s a function of supply and demand. Gold is used for jewelry. Art is in very limited supply and a way to show off one’s wealth. People do not buy bitcoins for some intrinsic value. They buy it for the hope that others would buy at a higher value; the definition of the greater fool theory.

Myth #6: Bitcoin is not a bubble

Of course it’s not. Even the Loch Ness Monster thinks it is not!

The creators of bitcoins have done something truly impressive: realize a longstanding dream of decentralized currency with software. But cryptocurrencies have become a tool for speculation. And their derivative, initial coin offerings (ICOs), have become a tool for scammers. As retired Americans start contemplating placing their life savings into bitcoins, the tech industry has a responsibility to call out the BS and stop brushing all criticism under the rug.

Ask an economist what the Greek financial crisis of 2010 and the fall of Rome in 476 AD have in common, and the answer is decadence: the luxurious self-indulgence that precedes a society’s decay.

Startups today have reached a new level of decadence unseen in the first tech bubble. What started as free lunch has evolved to five star meals by renowned chefs. $5 kombucha in the fridge? Check. An annual company trip to work remotely in an exotic location? Check. Want to experience burning man without the dessert heat, HR recruiters at one of the hottest startups will boast about their life-changing summer camp. Decadence, backed by a frothy late stage financing environment, has become the answer to all startup woes.

In their defense, startups have a hard time today. In the first tech bubble, a few would quit a startup to join IBM. And Microsoft required a move all the way up to Seattle. Today, Facebook or Google offer far more compelling financial and career opportunities than their old counterparts. They are also nimbler, smarter, and not only will they steal your talent but they will also steal your product ideas and out-execute you. Startups today have it rough!

But the answer is not decadence: no one quits their job because they don’t have a five star chef or a summer camp. Instead, employees quit when they do not like their boss or when they feel the place is a train wreck — a situation that decadence only amplifies. What employees want is a vision and purpose, a boss they respect, and a sense of reward for their efforts. Building, not indulging, is what keeps teams motivated.

This is not a case against having fun at the workplace. It is a case against throwing money at the problem of morale — only to realize decadence is a bottomless pit that will only make your problem worse.

Creating a purpose and building competent management are hard things to do — much harder than a company weekend in Cabo. But if it wants to succeed, a startup should invest its energy and capital there instead. Or succumb to the fates of Kozmo, Webvan, and Enron.

It was summer of 1999. I was four weeks into my dream internship: a software development engineer at Microsoft porting games to the Dreamcast game console. Bill Gates, still CEO at the time, had invited all the interns for brunch at his house — a sprawling mansion on lake Washington that featured medieval tapestries and AI-driven modern art. After a brief talk he opened the floor for questions. “When was the last time you have written code” I asked? “It must have been over ten years” he answered. Then he proceeded to say he actualled missed programming. But, as CEO, he needed the discipline do the things he didn’t necessarily enjoy.

I didn’t think much of his comment at the time. But today I realize it’s perhaps the most universal advice one can offer to a startup founder: Stop doing what you are good at and start doing what the business needs.

Most founders are good at and enjoy a particular function — whether it’s engineering or product or sales. Our strength becomes the hammer we use against every problem. If you’re a product driven CEO, you opt for developing version 2.0 instead of making the effort to put version 1.0 in front of customers. If you are a sales driven CEO, you keep selling the vision instead of focusing your energy into the painful minutiae of building a product.

This habit is exacerpated by the fairy-tale narrative our media loves to tell about founders: google founders programmed thier way into building a half a trillion dollar company. Airbnb succeeded becuase the founders loved great design. It’s a great narrative. It’s often a flat out lie.

Google built a great engineering culture, but it was their ad business innovations that made the company into the great business it is today. Airbnb didn’t succeed because of the founder’s passion for design; It succeeded because of the hard and painful work they put into building a two sided market place. It could have looked like Craigslist and still been quite successful.

Bill Gates was an amazing programmer infamous for writing a full basic interpreter that used under 4000 bytes of memory. But when IBM wanted an operating system, he knew he was better off licensing existing software instead of writing a new one from scratch. The nerdy programmer had to abandon the computer lab in favor of striking business development and licensing deals. And that’s how Microsoft was born.

Doing what you don’t enjoy is not easy! You must have the intellectual honesty to acknowledge that what you are good at is not what the business needs. And get up every day and do these hard things — from recruiting to fund raising and sales or marketing. It’s the hard way to build a startup. But it’s also the only way.

“We are killing it”, he said. I had asked what the company does as opposed to how well. “We are totally killing it.” He said again — eyes beaming with confidence. “My company is like Airbnb but better.” He proceeded to explain. They had grown 300% from the previous year. They needed no money. Sequoia were interested.

A year later they were out of business.

Spend a day reading online advice and one thing becomes clear: entrepreneurs are encouraged to be cocky: Fake it till you make it. Name drop. Confidence is the hallmark of great entrepreneurs. Steve Jobs was arrogant. The list goes on.

And it should not come as a surprise. While throughout history humility was considered a quality to cherish, we in America have taken a separate course in recent years declaring confidence and high self esteem as traits to desire— starting with Ralph Emerson’s “On Self-Reliance.” The Startup literature is a mere reflection of our values as a society.

But what if, for a moment, we question hubris. What if we give examples of entrepreneurs who were humble — who sold a vision without bragging about fake accomplishments, and who referred to their employees as the ones who actually built the company.

This is not some new-age call to disregard your self-worth. Being humble doesn’t mean to think less of yourself but rather to think of yourself less. And to think of others more.

I remember listening to Ben Silbermann, the founder of Pinterest, pitch the company in 2011 when it had five employees and millions of user. He never bragged about the rapid growth they were going through. He genuinely was concerned about the impact of scale on the community (and on the servers). He spoke of our company not my company. His humility served him well.

Being humble will make you a better listener. It will make you question vanity metrics and celebrate meaningful ones. It will lend your story credibility. And, if you’re one of the 99.99% of entrepreneurs who are not Mark Zuckerberg and who will likely face extreme hardship on the road to success, being humble will ensure your team and your investors follow you when times are tough as they did when times were good

Humility isn’t a trait you are born with. It’s something you practice. So in the middle of reading all the blog posts about self-discovery and self-improvement and self-help, try to replace your self-awareness with self-control. Your startup may be better for it. And you certainly will be.

You are about to present to your dream VC. The metrics look good but not great. Actually, they are not that good — they look like every other startup metrics in the valley. But, you are tempted. Did Mark Twain not quip that there are lies, damned lies, and statistics? So why not push the enevelope a little.

You can — and it happens every day. But you really shouldn’t. Because, as I will explain, there is Karma in reporting a company’s metrics: It all comes back and haunts you. And the more you exaggerate your metrics during fund raising, the more likely your startup will fail as a result — Just more spectacularly.

1. Underestimating your CAC

Your Customer Acquisition Cost (CAC) looks, well, like everyone else’s CAC in the category. It is costing $2-3 to get a click from Google, and, with a solid 2–3% conversion, your CAC is somewhere in the $66-$150 range.

The most common approach is to report the optimistic story: Take your cheapest keyword and your best performing conversion metric, discount your marketing spend a little bit, and end up with a healthy CAC that’s sub $50.

Or, you can go all out: just divide your total marketing spend by the total number of new users in a given period and, all of a sudden, your CAC is a super healthy sub $20 per user (Notwithstanding that techcrunch article that generated half your users). Surprisingly, many investors will let it go. People like good news.

But don’t. Eventually you will report on this CAC in a board meeting. Investors will dive deep into your acquisition cost by channel. And that super low number will make you lose a lot of credibility. You will also have wasted a bunch of marketing dollars in the process.

2. Overestimating your Active users

Active users is a bit of an outlier among metrics: it is both very important and functionally meaningless. No one can define it for your application. Only you can. (Personally I chose to define an active user as one who genuinely engages with the product without aggressive re-engagement tactics.)

The simplest way to exaggerate active users is to define it as anyone who used the application. This works particularly well during periods of rapid growth when every new sign up, by definition, is an engaged user. If you happen to be raising money during a time of growth, your engagement numbers may just look stellar.

Founders often try to game the system further by either accelerating the marketing spend before a fund raise, or by pushing for a very aggressive re-engagement tactics via email and push notifications.

Again, don’t. Eventually you will need to report on the actual metric. And what could be an improving engagement will now look stale or declining.

3. Only reporting the interval with the hockey stick

Every startup goes through ups and downs. As a presenter, you get to pick a sliding window from your startup’s lifespan to report on. Investors don’t expect you to pick the worst performing window, but they do expect you to report on a window that is both current and historically relevant. So if your startup’s growth has had an S curve shaped trajectory, don’t conveniently trim the timeline at the point of peek growth. An experienced eye will ask uncomfortable questions. A lesser experiened eye will let it pass. But will notice the numbers a few months post invesment and feel tricked.

4. Not communicating the bad metrics

This is at tricky one. So you have cracked the nut on virality but not on engagement. Or the Life Time Value of a user is through the roof but your CAC is also absurdly high. Why not just report the positive news.

Telling half a story is not techncially lying. But it is misleading. And research shows that while most humans don’t like to lie, most don’t mind when others are mislead. But you should hold yourself to a higher standard. Tell the full story to the extent you can. Investors get less excited about most deal a few months in anyway. Don’t make yours stand out by being labled the one with the lying CEO.

Why this is all bad

With enough experience, you will eventually realize that there is no game. There are real companies and there are houses of cards. Don’t think fake it till you make it. If the business is bad, no matter how you sugar coat your numbers you will still fail — just more spectacularly.

What you should do instead

Present a compelling vision. Have a genuine plan for improving your metrics that you need funding to execute on. Build a great team. Tell a story that you believe, not one that you want the world to believe. If the stars are aligned you will succeed in raising a round. And if the numbers are not great but the investors can sense your integrity, they will genuinely ask you to come back with a better business. That may be the best outcome for you after all.

There is no place for me in trump’s America: I immigrated to the US 21 years ago from Egypt, one of the many countries that falls under his various bans. But I came to this country with dreams — not just of economic prosperity but also of freedom and democracy. And i believe the recent wave of vilification of trump supporters in Silicon Valley, and the calls to boycott them, is a threat to this very democracy I have aspired to be part of.

Ever since the McCarthy era, political views have seldom been a threat to people’s professional careers in the US. As a result, I believe we have forgotten how dangerous this is. Look outside the US though and the world is full of examples for what happens when political stands can impact one’s career: political opinions will be pushed underground or publicly prosecuted (Middle East), we may create politics-based cronyism where supporters of one candidate or party are more favored for employment (China), or we may get a new wave of McCarthyism (USA) that kills the very essence of freedom of expression that has so drawn me to this country. At a minimum, Hilary supporters who live not in San Francisco but rather in the 43% of the nation that is Trump country will be forced to keep their opinions to themselves for fear of professional retaliation. That would be a shame.

Many will point out that Trump’s rhetoric is a threat to our democracy. That is indeed true. But our strength as a nation comes from abiding by our principles even with those who seek to destroy them. Paramount among these principles is freedom of expression.

In two weeks the American people will decide the next president. By most accounts, trump will lose and our democratic values will remain intact. If we keep vilifying those who support him, however, some of his ideas will have already won.

It was a peculiar sight. I had been in Cuba for 2 days without seeing a single smartphone. There it was: the public Wifi hotspot. A very public hotspot. Dozens huddled together in the heat glaring at their screens — some sitting on benches, others on the ground: Cuban teenagers browsing youtube on their Samsungs; Canadian tourists checking work mail on their iPhones; and older Cubans video conferencing with relatives — a cigar in hand. Many had paid a small fortune for the access card — $1, or the equivalent of a 2 day salary. And it was temporary: the access card only lasts for 30 minutes. That’s the state of internet in Cuba today.

Ever since the Internet was invented, Cubans were left in the dark. Fidel had judged correctly that opening up information would pose a threat to revolutionary ideas — already under threat during the Período Especial, the Cuban economic crisis of the 90s that followed the fall of the Soviet Union. Ask Cubans however, and many point a finger to El bloqueo, the US Embargo in place since 1960. My initial reaction to this opinion was skepticism. But a quick look at a map of underground caribean cables shows a tangle web of connectivity that bypasses Cuba. That the only cable reaching the country, the ALBA-1, was financed by the Venezuelan government further validates their point. It was a common theme: debating whether Fidel vs el bloqueo is to blame for the suffering of Cubans, the only certainty their suffering.

In July of this year, it all changed. In a surprising move, the government opened up the Internet to everyone — albeit with conditions. In addition to the hefty price tag it requires an identification card — the government official wrote down my passport number against the access card number; it’s only available via WiFi (Thanks to being late to the party, Cubans bypassed the cable phase and will never have to deal with Comcast); And it’s only accessible in public squares.

When I asked for directions to the office of Etesca, the state telecommunication operator and the only distributor of access cards, I was told a ten minute walk and an hour and a half wait. It was not a joke. Outside the office, a couple dozen Cubans were waiting In their infamous queues. Looking like a mess, the queue actually had a strict social order to it: Whenever a new person would show up, they politely ask for El ultimo, the person currently last in the line, then take the Ultimo spot. For 60 years Cubans have stood in these lines to buy anything — from basic cooking supplies to the occasional piece of pork or chicken. They did not seem to mind standing in line to get access to a glimpse of the outside world.

Availability was spotty: the first day I bought 2hour access in Havana, the next day they were totally out. In Trinidad supplies were running low so cards were rationed to 1 hour per buyer. Outside, scalpers were selling the cards for a 100% premium.

Whatever Internet they got was surprisingly fast — I measured about 1 Mbps for downloads and uploads. The constraint on time and place creates interesting habits. Some practices — writing an email offline then sending it once connected — are reminiscent of the early internet days. Others are absent: when I asked about dating apps I got back a rather confused look. And some practices around conferencing are truly novel. In many instances, multiple members of a family were huddled around one small screen to video conference with distant relatives. In one instance, a father was video conferencing while his teenage son was watching a muted youtube video in an adjacent window. Their card ran out after a few minutes. The father sighed in frustration. It was painful to watch.

But perhaps what’s truly uniqe about internet access in Cuba is the public nature of it all. At a time when the Internet is associated with social isolation, Cuba is a sweet reminder of a time when the Internet was exactly the opposite: a means for brining people together. Going to Cuba is often described as a trip to a different time more than place. The Cuban Internet is no exception.