Jean-Michel Bouchard

@karrmannmatt @theo @ThePedroProenca @mil000 Actually I stand corrected on Amazon, I was wrong there.

Confirming the meaning of cliff at fb. “Fully vested” never happens due to refreshers. The only cliff at fb is when your original grant is fully vested. AFAIK Google and Amazon are the same RSUs are granted every 3 months. Yes, from your first quarter on. Theo is right in most cases. The cliff *usually* means when you can exercise your options / get your first non liquid RSUs at startups and medium tech. TikTok / Bytedance followed (still does I assume?) that rule.

? The “Cliff” usually refers to the point where you go from 0 equity to actual equity. Industry standard is 1 year. OAI follows that afaik. From that point, you vest on a schedule. Every month or every 6 months. Never seen a continued vest schedule that is yearly, but it wouldn’t be a “cliff” either way.

@aakashgupta the person that decides WHAT to build is the person that starts building. PM, engineer, SDR, who cares

Andrew Chen just called PM the most important role in tech again. His argument in one line: when anyone can build, the person who decides WHAT to build becomes the bottleneck. Boris Cherny’s Claude Code team at Anthropic shows what this looks like in practice. They ship hundreds of prototypes before committing to a feature. Boris personally runs 5 parallel Claude instances and ships 20-30 PRs a day. The team built Cowork, a full product for non-engineers, in about 10 days. Productivity per engineer grew 70% even as Anthropic tripled headcount. The cost of building dropped through the floor. The cost of picking the right thing didn’t. Anthropic’s PMs don’t even write traditional PRDs anymore. They review working software at 9am, kill 80% of it by noon, ship the rest by end of week. Pattern-matching across user research and technical feasibility while staring at a working prototype. This is why AI PM offers at OpenAI, Anthropic, and Google DeepMind now run past $1M total comp. Anyone can build. Almost nobody can look at 15 working prototypes and call the 3 worth scaling. That judgment didn’t matter as much when shipping took 6 weeks. It matters enormously when shipping takes 45 minutes. Andrew nailed the diagnosis. The harder question for every PM: can I be the one who picks? Here’s how to build the muscle: 1. Modern PRD guide: news.aakashg.com/p/product-requ… 2. AI Prototyping tutorial: news.aakashg.com/p/ai-prototypi… 3. AI Roadmap: news.aakashg.com/p/ai-roadmap 4. PM Operating System: news.aakashg.com/p/pm-os 5. Learn this live in my cohort: landpmjob.com The bottleneck moved. The pay moved with it.

@siddontang That’s a good take but that agent runtime seems like a consistency nightmare. I’ll assume that it’s designed with CQRS patterns in mind or this will create a new level of slop.

@vitaliidodonov “Trust compounds faster than revenue.” No better words to hear in the agentic era.

Here’s my backstory on how we took Stan Store to $30m ARR: When my co-founder and I started Stan, it was just the immigrant grind for years with lots of frugality and sheer hustle. There was even a time when we returned an IKEA table after using it in an Airbnb for some time, asking ourselves: "Are we building something that looks like a startup? Or something that actually helps people make a living?" And that question drove everything that followed. Early on in our journey to $30M ARR, we made a decision most advisors disagreed with. - No transaction fees to use Stan - Flat subscription, no upsells - $29/month price. People told us we were leaving millions on the table by not charging a transaction fee. Like every single one of our competitors did. And guess what? They were right. We left the money there on purpose. Because we realized that the moment you take a cut, you’re no longer on the Creator’s side. You’re just the middleman. We wanted Creators to win more because they used Stan. Not win less because we skimmed the top. Trust compounds faster than revenue. That turned out to be true. As growth picked up, so did the pressure. “Hire faster”, “Build a big team”, “Spend to show momentum.” We ignored all of this. Instead, we asked one question over and over: "What breaks if we don’t hire?" Most of the time, the answer wasn’t “people” … but systems. So we instead, we automated and simplified as much as we could. Deleting features. Building self-serve tools. Then, and only then, we hired people. Today, Stan serves 80,000+ Creators: They make $200M+ per year using the platform. Three years into building, we celebrated: ~$30M ARR. ~30 people. ~7 engineers. And a relentless focus on: – trust – speed – product – and not doing work that doesn’t need to exist So when I see posts that talk about Stan as the “fastest-growing startup I’ve seen”… I get it. But the real story that needs to be told is not the one about speed. It’s the one about restraint. We said no more than we said yes and we didn’t optimize for looking successful. What we optimized for was being helpful and creating fundamental value for our customers. We like to say that we didn't build Stan alone. The Creators built it with us. One at a time. Over the years. And that, out of any ARR, MRR or GMV… That’s the only leaderboard that actually matters.

@threepointone let’s go! built in rest time will allow you to ship even faster when you’re back

setup OoO responder for the first time in my life cheeky airport pint and I’m outta here

@KVisner @simonbrown Does it work? Does it provide more value than it cost? 🚢

I think everyone needs to go back to the KISS principle. It seems like, instead of creating the minimal amount of code to get the job done, everyone is letting AIs create maximal solutions to even simple problems, and then leaning into it by drafting maximalist specs. Can you really reason about how the project should look in a couple of hours after the AI has generated 100k lines of code? I bet not.

Spec-driven development ... the 1990's called and they want their processes back. As a junior developer in the late 1990's, before writing code, I was asked to: - document my understanding of the feature I'd been asked to build - document the code/DB schemas I was planning to add/modify/remove - document the tests I was planning to run afterwards This would be iterated upon a few times, and finally I was permitted to write code. This post on Reddit amuses me... 😄

Referral Recruiting is probably the biggest founder life hack.

@simonbrown ✨evidence driven development ✨

@vaibcode that’s a scary thought, control system theory will help to stabilize this harness

@ohnojmb basically. token factory is alive and well.

Being at the "frontier" is just about stacking while loops and "intelligence" is just a measure of how long while true can run uninterrupted. on a separate note, this is also a good way to measure human intelligence.

@rvcas what happened to rust/ts/zig/sql? 🤣

Odin lang genuinely looks pretty nice

@shellscape @karrisaarinen platform folks deserve love too 🥲

@karrisaarinen I do not understand sending feels to the freaking massive tech behemoth that is Microsoft. They have all the money. They can get all the people. They prioritized the wrong things and the platform suffered. hug ops dismisses accountability.

Hug ops to GitHub. They’re seeing unprecedented growth in a very short period of time. Much of it is probably coming from free accounts that don’t generate revenue. They’re essentially providing a lot of the infrastructure that coding agents depend on completely for free or for very low price. Think about your token spend vs GitHub spend.

@aakashgupta Computational Psychiatry; it’s a thing and it needs more research !

The blue figure is experiencing an amygdala hijack. The red figure is experiencing something worse, and almost nobody has a famous name for it. Daniel Goleman put "amygdala hijack" in a bestseller, so the left state has cultural shorthand. The amygdala fires in roughly 12 milliseconds, releases cortisol and norepinephrine, and the prefrontal cortex goes offline before your conscious mind has even registered what happened. Acute phase lasts 20 to 60 minutes, which is how long stress hormones take to clear the bloodstream. You don't choose to stop thinking. The biology pulls the plug. The right figure is the default mode network in overdrive. The same network that runs planning and autobiographical memory, looping on itself until it produces rumination. Susan Nolen-Hoeksema's longitudinal studies showed ruminative response style is one of the strongest behavioral predictors of major depression onset we have. No bestseller. No catchphrase. More common, more dangerous, harder to spot from the inside. Two panels, same axis, opposite polarities. Hyperarousal on the left, hypoarousal on the right. Dan Siegel called the middle the "window of tolerance." Founders and operators oscillate between these states for years and call it productivity. Crisis mode for the launch. Spreadsheet mode for the recovery. The crash on the left is loud (you yelled in a meeting). The crash on the right is quiet (you optimized a strategy nobody asked for over six weekends). The intervention is the same on both sides. Vagal tone. The vagus nerve is the physical wire connecting feeling and thinking, and you measure its function with heart rate variability. Breathing at 5.5 to 6 breaths per minute synchronizes heart rate with respiration, maximizes HRV, and pulls you back into the window. Paul Lehrer's lab has 20 years of trials behind this. The effect shows up in the first session. Cold water on the face does it through the mammalian dive reflex. Humming does it because the vagus nerve runs past the vocal cords. The comic is funny because it's accurate. Feeling and thinking share a power supply. The supply is the vagus nerve. Most of us spent decades learning to override it.

@aninibread @CloudflareDev this is work for my agent. api pls

small change, big quality of life: you can now empty @CloudflareDev R2 buckets and delete folders straight from the dashboard. cycling out old backups, clearing millions of temp files, resetting a dev environment, just click the button and go enjoy your day. 🗑️

We are at a point where the most important skill for a software engineer is “Evidence Driven Development”. claude/codex : “do X, show me the evidence”

If you've wanted to get into open source and have interest in JSX, email, ASTs, transforms, and plugins, I'd love to have you on board with jsx-email. More than happy to help get you going if you've never worked with open source before.

@rvcas Don’t forget SQL

The final 4 languages are TypeScript, Go, Rust, and Zig