Jon Crowell

right now everything in the world is telling you to go faster, ship more, add that feature, start another project so i'm actively working on feeling ok not doing any of that

@willsentance I would love to see a few examples of your full auto research process. How are you setting it up? What problems are you solving? What questions are you answering? What processes are you improving? And would love to see the evals breakdown and how it impacts all of the above.

autoresearch is getting bigger. automatically improving prompts, workflows, agentic outcomes. the hardest part is devising the metric goals that autoresearch tunes towards. in ML, its easy. for all other agentic workflows, you have to get creative. this is when evals come in. evals dont just measure the outcome, they also measure how the agent got there. reasoning, tool calls, hiccups and assumptions. even if you arent autoresearching.. evals take your agentic workflow to the next level, easily eclipsing the use of unit tests and spec checklists. who would benefit from a breakdown of evals?

- Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - Wow, feeling great, it’s so convincing! - Fun idea let’s ask it to argue the opposite. - LLM demolishes the entire argument and convinces me that the opposite is in fact true. - lol The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.

All to the tune of Flowers Are Red, by Harry Chapin.

🇧🇷🗣️ | Ex Brazil midfielder Felipe Melo on Brazil’s current situation: “Brazil has great players but they have the wrong attitude. If anyone comes on here and tells me Brazil don’t need Neymar I’ll block them. Neymar with ONE leg is Brazil’s no.10.”

This Brazil team is hard to watch.

loved her work but this is really disappointing. i hope she can clear this up

If the AI revolution doesn’t result in 10x higher quality software when being driven by experienced engineers, I’m gonna be so disappointed in us. We should be “over cooking” the things we undercooked before because of the cost.

I’m stuck here with several thousand people at the Houston airport. The wait times just go get through security are up to four hours. … Political leaders— get on with it. Put guardrails on ICE and pay these TSA workers. It’s not that complicated.

We’re at a point in history—not nearing it, but here—where everyone is going to have to decide if we are content to numb ourselves with an endless stream of fentanyl-like digital slop or if we are going to fight for our humanity and touch grass and challenge ourselves and create and contribute and love. Lawsuits like this are setting an interesting precedent. But no one is coming to save you, at least not any time soon. Protecting your attention is a radical act. Don’t let the internet turn your brain turn to sawdust. My decade-plus researching and reporting for The Way of Excellence taught me that people are most satisfied and fulfilled when they care deeply about meaningful projects. When they have mastery and mattering. When they do good work and love good people. Nobody told me they feel or perform their best when they are mindlessly scrolling. The greatest risk of the modern world is we go wherever the current takes us, like automatons floating along an algorithmic conveyor belt. The only thing that separates us from this dystopia is ourselves. Our agency—our attention, our capacity to think, create, and love—must be fought for. Read a book. Play a sport. Make art. Garden. Go on a date. Coach a team. Meet new people. Have IRL conversations. Don’t be a zombie. Do cool shit. Live.

the generation that refused to accept cookies. is now giving AI access to their desktops, files, and bank accounts.

I was applying for Australian citizenship and the interviewer asked, "Do you have a criminal record?" I said, "No. Is that still required?"

Good tip for avoiding cognitive debt in codebases where AI has run wild: Design the interface, delegate the implementation

@anishmoonka What research are you referring to? Do you have a link?

The research behind this is wild. Your brain can’t flip from full alert to sleep like a light switch. It needs a runway. And reading builds it faster than almost anything else. A University of Sussex study found that just 6 minutes of reading cut stress by 68%, more than music (61%), tea (54%), walking (42%), or video games (21%). The effect is surprisingly physical. When you read, your nervous system shifts from fight-or-flight into rest-and-digest mode. Heart rate slows, breathing deepens, muscles release tension. The neuropsychologist who ran the study, Dr. David Lewis, described it as entering “an altered state of consciousness,” where focused imagination activates the part of your brain that tells your stress response to stand down. A 2021 randomized trial tested this directly. Researchers split nearly 1,000 people into two groups: read a book in bed for seven nights, or don’t. After one week, 42% of readers reported better sleep versus 28% of non-readers. Nothing else changed. Now compare that with what 86% of Americans actually do before bed: scroll their phones for an average of 38 minutes a night. A 2025 Norwegian study of 45,000 university students found that every additional hour of screen time in bed raised insomnia risk by 59% and cut sleep by 24 minutes. A separate American Cancer Society study of 122,000 adults found daily screen use before bed was tied to 50 fewer minutes of sleep per week. Screens hit you with two sleep-blockers at once. Blue light suppresses melatonin, the hormone that tells your body it’s time to sleep, by about 50% according to a Harvard study. But the bigger problem is the content itself. News, social media, work emails, all of it fires up your brain’s threat-detection mode and spikes your stress hormones right when they’re supposed to be at their lowest point of the day. A physical book sidesteps both problems entirely. The long game matters too. A Yale study tracked 3,635 adults over 12 years and found that people who read 3.5+ hours per week were 23% less likely to die during the study. That worked out to living roughly 2 years longer, regardless of gender, wealth, or education. Books beat newspapers and magazines. The researchers pointed to deep, sustained reading creating a kind of workout for the brain that protects it as it ages. So the 5-10 minutes he’s describing? The science says 6 minutes is the threshold where your body starts winding down. His brain is switching off its stress response and easing into a state where sleep becomes almost automatic.

The "10x AI Developer" is a MASSIVE lie. You are just a 1x Developer generating 10x the technical debt. The entire tech industry is high on the illusion of "vibe coding" right now. The popular consensus is that because Claude and Devin can spin up a backend in 45 seconds, software is now infinitely cheaper to build. Here is the provocative reality nobody is budgeting for: AI is about to make software engineering significantly MORE expensive. Everyone is cheering for code generation, but completely ignoring the Verification Tax. When an AI agent writes 5,000 lines of code, it is optimizing to pass the immediate test. It is not optimizing for human readability. It relies on brute-force loops, repetitive logic, and bizarre architectural shortcuts that just happen to compile. Fast forward 12 months. Your business needs to pivot, or a core dependency breaks. You are now staring at a 50,000-line black box that no human being actually wrote, understands, or can safely modify. You cannot simply "prompt" your way out of architectural collapse. When the machine-generated spaghetti finally breaks, you won't be saved by a $20/month LLM subscription. You will have to hire a top-tier Principal Engineer at absolute premium rates just to untangle the mess your "autonomous swarm" created. We are treating code generation as a pure productivity win, but code is a liability, not an asset. Stop measuring how fast your team can generate syntax. Start measuring how quickly they can debug it.

Dirty industry secret: nobody really knows the best way to use AI for software development 📢 - 24 months ago, we copy-pasted from ChatGPT - 18 months ago, we jumped between ask mode and agent mode - 12 months ago, we told AI "you are a senior developer" - 9 months ago, we built MCPs - 6 months ago, we switched to plan mode - today, we're obsessed with skills All of this (and much more) are just early experiments and temporary hacks in a very young and quickly evolving field. So when someone says their workflow is the optimal one, they're confused at best. Stay curious, stay open, stay in control and stick to the fundamentals, and you'll come on top in this amazing tech revolution. Enjoy the ride!

During an interview in 1974, Corrie ten Boom shared how she once encouraged fellow believers in Africa with one of her father's memorable lessons on why Christians need not fear being strong enough to endure suffering: "I once said to my father (I was still a little girl), 'Daddy, I will never be strong enough to suffer for Jesus.' And Father said, 'When you go to travel with a train to Amsterdam, when do I give you the train ticket? Three weeks before?' I said, 'No, Daddy, the day that I go to travel.' And Father said, 'That's what God does. Today, you do not need to have strength to suffer for Jesus, but the moment you will have the honor to suffer for Him, He will give you all the strength.' And then I was confident. And I said to these people, 'When you have to suffer for Jesus, the Lord will give you the train ticket.'"

Asked the AI to make a fun slide about potential de-skilling from AI use. Need to specify more clearly what I mean by "fun".

PM openings are at the highest levels since 2022 There are over 7,300 open PM roles at tech companies globally, and trending up. This is 75% above the low we saw in early 2023, and already up nearly 20% since the start of this year. Today we have the most open PM roles we’ve seen in 3+ years.

Here is a little trick: I don't use a TODO app. I use a pinned Codex thread for my long-running TODO. The initial system prompt is to create a master TODO.md and keep it updated with items added by date, etc. It's natural intelligence for my tasks, and it's been very useful so far.