← Back to Writing

Existential psychosis: what building with the latest models feels like

March 20, 2026 · 2 min read

agentsbuildingaxonlinkedin

I don't know how to explain what building with 4.6 and 5.4 feels like without sounding crazy.

You sit down, you brief your agents, you delegate across branches, research, engineering, finance, validation, all running in parallel, all actually doing the work. Not summarizing. Not suggesting. Doing.

And every single time the output lands you end up just staring at the screen.

Existential psychosis. That's the only way I can describe it. A kind of mental shock that doesn't go away no matter how many sessions you run. If anything it gets worse the deeper you go.

Here's a concrete example of what I mean.

I'm currently preparing 15 client engagements simultaneously. Research branch running full industry and market analysis for each company. Finance branch identifying the specific revenue opportunity and ROI case for every single one. Engineering branch building a BI dashboard to give us a structural competitive moat. All three running in parallel. Each branch producing real, usable output, pitch strategy, financial opportunity mapping, AI implementation roadmap, industry by industry, company by company.

That is not a chatbot. That is an operation.

And the part that really breaks your brain is what happens after.

Every deliverable gets checked by QA agents before it moves forward. Then a validation layer reviews the full output across branches. Then everything gets logged, agent by agent, session by session, so the system can identify where gaps are, where specific agents need to level up, and where the model itself has room to improve. The self-improvement loop is built in. Every run the ceiling moves up.

We are talking about going from planning to strategy to finance to implementation to QA to validation to improvement logging, fully orchestrated, fully parallel, fully compounding.

I've been building for a while. Nothing has been this fun. Nothing has felt this close to something real.

If you've gone deep with the latest models you already know. If you haven't, go deeper. The shock is worth it.

Originally posted on LinkedIn.

Recent Posts

Writing
Coding
Career
About