Here I can
OpenAI is deliberately stacking powers: funding and coordinating the physical layer, converting that into a trusted, best-answer experience, and binding it together with identity, memory, and apps—run by a product machine.
If you're going to really exploit the capabilities of these new tools, you need to be operating at the top of your game. You're not just responsible for writing the code - you're researching approaches, deciding on high-level architecture, writing specifications, defining success criteria, designing agentic loops, planning QA, managing a growing army of weird digital interns who will absolutely cheat if you give them a chance, and spending so much time on code review.
…one of the biggest differences I've seen between people that really know how to build agentic workflows compared to people that are less effective at it is the ability to drive a disciplined development process, specifically one focused on evals and error analysis.
The unit economics of SaaS create a predictable organizational structure: ARR compounds through new bookings and net retention, so companies optimize ruthlessly around those metrics. This means the roadmap becomes a sales tool—every feature maps to a deal or a churn save. The best orgs maintain product integrity while doing this. The median org ships the top 10 feature requests from their biggest customers and calls it a strategy. From Ben Thompson's excellent interview with Bret Taylor:
Broadly speaking, sales and marketing matter a lot in enterprise software companies and tend to be the gravitational center... Because at the end of the day, annual recurring revenue is an annuity, it throws off cash every year, you basically need to add to that annuity by signing new recurring revenue contracts or adding to them, and you subtract attrition. That's just how software as a service works.
With that, when you stop growing, the business model breaks down. That's where things like the Rule of 40 come from, where your growth rate and your EBITDA margin sort of need to be in lockstep. You can grow very profitable with that business, when you stop growing, no one wants to work at one of those companies.
As a consequence, you end up with a really customer-centric, go-to-market-centric orientation, and product serves that. The best enterprise software companies, that voice of the customer dictates their product roadmap, and they can really meet that demand.
The worst ones stop innovating on the product and hold your feet over the fire in the sales process. Depending your level of cynicism, all enterprise software companies have been guilty of each at different points.
Another interesting thing about this is how the business model shapes culture. When you're growing 80%, product can lead and sales follows. When you're at 15%, sales leads and product follows. It's hard—the companies that stay interesting are the ones that maintain product velocity even as they scale GTM.
Living off borrowed time, the clock tick faster.
Anthropic is riffing on Madvillainy and I'm here for it.
I used to think network effects would win out, and we'd end up with just one or two stables, but I don't believe that anymore. Cross chain swaps are getting more efficient by the day, as is intra-chain inter-stablecoin swapping. In a year or two, I think many intermediaries in crypto will just show your deposits as generic "dollars" or "dollar tokens" (rather than USDC or USDT) and they'll guarantee you redeemability in a stablecoin of your choice.
Already we are seeing this with a lot of the fintechs and neobanks that are product rather than crypto first. They care most about UX and would rather offer clients the best experience instead of honoring crypto traditions. So they just show your balance in USD and manage the reserves on the back end.
Three major AI video launches in three weeks, and each company picked a radically different target audience:
...compelling AI video generation is very much here, and it's widespread: over the last three weeks we have actual AI video products from Google, Meta, and now OpenAI.
...it feels like each company has an entirely different target audience: YouTube is making tools for creators, Meta is building the ultimate lean back dream-like experience, and OpenAI is making an app that is, in my estimation, the easiest for normal people to use.
In this new competition, I prefer the Meta experience, by a significant margin, and the reason why goes back to one of the oldest axioms in technology: the 90/9/1 rule.
- 90% of users consume
- 9% of users edit/distribute
- 1% of users create
If you were to categorize the target market of these three AI video entrants, you might say that YouTube is focused on the 1% of creators; OpenAI is focused on the 9% of editors/distributors; Meta is focused on the 90% of users who consume.
...it's striking how this target market evaluation tracks with the companies themselves: YouTube has always prioritized creators, while OpenAI's business model is predicated on people actively using AI; it's Meta that has stayed focused on the silent majority that simply consumes...
- You learn more per unit time because you make contact with reality more frequently
- Going fast makes you focus on what's important; there's no time for bullshit
- "Slow is fake"
- A week is 2% of the year
- Time is the denominator
It has been clear for a long time now that ChatGPT is a powerful tool for commerce, and the company has been taking baby steps to close the loop, with things like product carousels, an in-app browser (which lets OpenAI capture web usage and attribution), and the overall concept of having a router deciding how to answer every query. This, however, is an explicit step into the transaction itself — and it's almost certainly not the last.
OpenAI partners with Stripe to launch instant checkout within ChatGPT.
…Meta can still build a good product. Vibes may not end up being for everyone — I hesitate to generalize too broadly from what I like, given my rather esoteric media consumption habits — but I think it is executed very well, with a strong and unique point of view. That sort of product sensibility is something that has generally been missing in AI.
..Meta has the courage and conviction to lean into what AI can uniquely do. There are a lot of things to wring one's hands about when it comes to Vibes, but Meta is leaving that hand-wringing to the peanut gallery, and actually shipping. There's something to be said for that.
“Scaling laws are real, ofc. ++ model power keeps raising the ceiling. But capability ≠ adoption. The moat in workflow AI is bottling institutional context and making it useful in the moment of work.”
…
“Pretrained models are powerful, but not turnkey. Extracting Quinn’s win rate, UCSF’s pathways, or Zappos’s ethos is closer to RL than pretraining. You shape behavior with constraints and feedback until it fits the environment. Workflow products do that shaping.”
…
“The Bitter Lesson is obvi right: scaling keeps raising the baseline. But raw capability isn’t adoption, because no user wants to give superintelligence full context all the time. App winners are the ones who bottle context into systems trusted at the point of action.”
2026 will be a pivotal year for the widespread integration of AI into the economy:
- Models will be able to autonomously work for full days (8 working hours) by mid-2026.
- At least one model will match the performance of human experts across many industries before the end of 2026.
- By the end of 2027, models will frequently outperform experts on many tasks.
Early GDPval results show that models can already take on some repetitive, well-specified tasks faster and at lower cost than experts. However, most jobs are more than just a collection of tasks that can be written down. GDPval highlights where AI can handle routine tasks so people can spend more time on the creative, judgment-heavy parts of work. When AI complements workers in this way it can translate into significant economic growth.
“In AI, we talk about ‘product overhang’, and this is what we discovered with the prototype. Product overhang means that a model is able to do a specific thing, but the product that the AI runs in isn’t built in a way that captures this capability. What I discovered about Claude exploring the filesystem was pure product overhang. The model could already do this, but there wasn’t a product built around this capability!”
…
“Building and testing 5-10 prototype ideas in a day is possible with AI agents. Prototyping used to be so time consuming that if there were two days to prototype, it was lucky to have two distinct prototypes built by the end. But now, agents can build prototypes very quickly, so tests of 5-10 prototypes per day are easily done, as the Claude Code team did.”
…
“Claude generates Markdown which is rendered in the terminal. Before the release, Anthropic engineers complained that nested lists don’t really look that good, and that the spacing between paragraphs was off. The problem was with the markdown renderer. Boris tried all the existing ones, but none looked good inside the terminal. The day before public release, Boris used Claude Code to vibe code a Markdown parser and renderer. And the result was better than anything preceding it, so they shipped it! Boris says he thinks no other terminal has the same Markdown rendering.”
I can’t overstate what a massive opportunity this is: every item in every YouTube video is well on its way to being a monetizable surface. Yes, that may sound dystopian when I put it so baldly, but if you think about it you can see the benefits; I’ve been watching a lot of home improvement videos lately, and it sure would be useful to be able to not just identify but helpfully have a link to buy a lot of the equipment I see, much of which is basically in the background because it’s not the point of the video. It won’t be long until YouTube has that inventory, which it could surface with an affiliate fee link, or make biddable for companies who want to reach primed customers. More generally, you can actually envision Google pulling this off: the company may have gotten off to a horrible start in the chatbot era, but the company has pulled itself together and is increasingly bringing its model and infrastructure leadership to bear, even as Meta has had to completely overhaul their AI approach after hitting a wall. I’m sure CEO Mark Zuckerberg will figure it out, but Google — surprise! — is the company actually shipping.
“…the company successfully navigated one paradigm shift, and is doing much better than I originally expected with this one. Larry Page and Sergey Brin famously weren’t particularly interested in business or in running a company; they just wanted to do cool things with computers in a college-like environment like they had at Stanford. That the company, nearly thirty years later, is still doing cool things with computers in a college-like environment may be maddening to analysts like me who want clarity and efficiency; it also may be the key to not just surviving but winning across multiple paradigms.”
Our whole thesis is that for every company in the world, their AI agent will be their most important digital interface they have, as important as their website or their mobile app. Just like mobile apps didn’t replace websites, and it also differs but not every retailer even has a mobile app. Some do, some don’t. You’ve covered this a lot, just like do you have the brand equity to warrant that?
I think across the board, your AI agent, whether it’s published in your mobile app or available via WhatsApp, or available over CarPlay or whatever it might be, that will be the primary digital interface to your company. We’re really focused on customer experience more broadly. What’s exciting for us, if we do our job, we’re the most important AI agent you make, we’re the one that actually powers your customer experience.
“Upon opening a project or folder, you’re likely to jump into editing files. This means Cursor needs to generate autocomplete suggestions, which the Cursor team calls tab suggestions. A low-latency sync engine powers the ‘tab model’. This generates suggestions that are greyed out and can be accepted by hitting the ‘Tab’ key. The suggestions need to be generated quickly in ideally less than a second. This ‘tab model’ must be as fast as possible, and data transfer as low as possible. There’s always a tradeoff between how much context to send, and the quality of the suggestions: the more relevant context Cursor can send, the better the suggestions. However, sending lots of context can slow down the display of suggestions, so getting this right is one challenge for Cursor’s engineers.”
I had this goal, which was we needed to somehow create a more imaginative world. I mean, one of the biggest risks in the world I think is a collapse in belief, a belief in ourselves, a belief in the future. And part of that I think comes from a lack of imagination, a lack of imagination of what we can be, lack of imagination of what the future can be. And so this imagination thing I think is an important pillar of something that we need in the world. And I was thinking about this and I saw this, I'm like, "I can turn this into a force that can expand the imagination of the human species." It was what we put on our company thing now. And that felt realistic. So that was really exciting.