Sean Hess

Matt is such a great thinker. When he joined Google, he wrote a strategy paper that was, by far, the clearest, most coherent thing I read while I was there. He articulated a clear Plausible Theory of Success (I wonder where he got that habit from? I'm looking at you Satya Nadella 🤣 ) If you aren't following Matt - fix that now.

111

12 Comments

We don’t talk enough about this kind of leadership. Last week, forgecode shipped a feature that degraded response quality. Instead of hiding it, they: - Admitted the mistake publicly - Explained what went wrong - Fixed it promptly In the AI space, it’s easy to chase growth at all costs. But accountability is what builds real trust. The companies that own their mistakes and put customers first? They’re the ones that will win long-term. We need more of this energy in tech.

24

I've caught myself three or four times building something where the AI enthusiastically helped me do something I shouldn't have been doing at all. Features nobody asked for. Content that was solving the wrong problem. Directions that felt productive because the agents were producing output but weren't actually getting me anywhere useful. It took stepping away and coming back each time to realize the direction was off. The agents never said "are you sure this is worth your time?" They just executed. Enthusiastically. I'm building a consultancy right now. Just me. AI tools handle what would have required contractors or employees a year ago. The site, the content pipeline, a whole automated research and drafting system, the operations scaffolding. I can move fast in a way that genuinely wasn't possible before. But your ideas, your biases, your blind spots go unchallenged when it's just you and a fleet of agents. The agents are good at executing. They don't push back on the concept. Just the implementation. That $401 million telehealth startup everyone's been sharing (one guy, $20K starting capital, hired his brother) is real. The economics of solo building have genuinely changed. But that story is an outlier being treated like a playbook. The build phase collapsed. That's worth celebrating. But building was always the easier part. Knowing what to build, who it's for, when to stop, whether it's actually working. That's still on you. Nobody automated judgment yet. IDK... maybe that's the point.

23

12 Comments

Over the last ~3 years, I’ve run ~83 AI-orchestrated software experiments. Some worked. Some failed. Some failed in ways that were more useful than success. I finally pulled the work together into a portfolio site that documents the patterns, case studies, and lessons I’ve learned while building with AI as the primary implementation layer. The site includes case studies on: - LLMs as runtime systems - hallucination as a pipeline problem - personal AI memory and context limits - mixing deterministic and non-deterministic systems - production agent stability - browser automation against difficult/adversarial targets - methodology for scaling AI-assisted development The main takeaway: applied AI engineering is not just prompting. It is architecture, constraint design, evaluation, debugging, product judgment, and knowing when not to let the model decide. Would love feedback from anyone building, hiring, or thinking seriously about applied AI systems. https://lnkd.in/gYcSdYEJ #AppliedAI #AIAgents #AIEngineering #Automation

4

Lime is a new parser generator similar to Yacc, Bison, ANTLR, etc. ad infinitum save one thing... it's faster and has the ability to merge grammars at runtime. See the example of a calculator that starts knowing + and - but then adds ^ for exponent, then adds ^ again for bitwise or. That can't work, right? https://lnkd.in/e9XbPcT7

2

2 Comments