The strangest part of working with AI agents all day isn't how much gets done. It's that the job now feels more like management than engineering, and the "soft" skills turned out to be the hard ones.
https://t.co/FlKJlAq8GE
James C Scott's *Seeing Like a State* is one of my all-time favorite books, and it influences a lot of my writing and day-to-day work. I wanted to pull out one of its best ideas, legibility, and show how AI agents improve and obscure work.
https://t.co/chG2qFENYi
I've talked a lot about life at Delphos Labs, and now I'm very happy to share a much deeper dive into our mission and vision. We have some really exciting announcements coming over the next few weeks too. I can't wait to share what we've created.
https://t.co/Tpr6RV8jWp
New post: I've seen lots of takes on software abundance, so I built a risk-return framework for the SDLC. Classify each phase by how costs distribute and benefits scale, then follow the math to where human time should go. Hope you enjoy it!
https://t.co/A8xNelj4Fp
Our automated reverse engineering systems are starting to find vulnerabilities. First of many.
All black box. Only the compiled binary.
Identified using code reasoning.
Responsible disclosure underway.
You can just do stuff is definitely my moto for 2026.
I recently decided to automate the newsletter workflow for my blog. Before, this would have been a weekend project. Now with agents, it took a couple of hours.
https://t.co/KioCtkdqB1
Why does AI coding feel like magic while AI doc writing feels like a chore?
I wrote about how we fix the "Knowledge Stack" to bring agentic workflows to everything from legal briefs to design docs.
https://t.co/1XdFhQgPvq
Working with LLMs consistently has revealed that context is at the heart of all of this, and it will remain critical to doing it well in the future.
I hope you enjoy the post.
https://t.co/mN7sZwJVrf
But since forecasting is hard, I wanted to take a different tack for this. I wrote this up to explain why we think so much labor might be automated by AI. My model is a simple loop.
I'm continuing to iterate on my website and blog. I thought adding a form for a newsletter might be interesting, and it's an good place to test case using agentic cools for coding.
I hope you enjoy it!
https://t.co/oRTRRks5kb
I've spent the last couple of weeks thinking more about engineering workflows and AI tools, motivated by the fact that Claude Code now integrates with IDEs. This continues to push forward the idea of context engineering and how it might impact your work. https://t.co/0vfALlb4ex
One of the things that I was most excited about after leaving Google was a chance to start some of my own projects again. To that end, I've rebooted the old blog. I wrote down some thoughts on how modern tools are reshaping writing software.
https://t.co/Yr1VPE5HMA
Machine Learning Meets Malware. If cognition becomes an API call and malware can be reverse-engineered by an LLM, then what’s left of “zero trust”?
Caleb Fenton joined @patio11 for a chat on AI, nation-states, and the new front in software security.
🎧https://t.co/FcExUUt6tz
I wrote a blog post with @jnvmiles and Ka Wong on better understanding the quality data generated by people, which is a big part of what we do at Google.
I'd love to hear what you think about it.
https://t.co/ePDF3BBr8F