Vexjoy

Typesafe.ai’s Jev returns typed decisions with probabilities instead of prose.

Typesafe.ai’s Jev replaces text generation with calibrated probability.

A new paper analyzed 61,608 AI-generated stories and found 93.2% detection accuracy using only narrative structure, no style analysis needed.

We pulled five rules from a 1956 Bertrand Russell essay, wired them into our agent, then spent eleven phrasings wording one clause about using fewer words.

A colleague asked why the agent routes to pipelines and workflows for me but not for him.

GSD’s creator rug-pulled and vanished.

When your team adopts AI, mistakes are inevitable and necessary.

I ran a Codex autonomous loop for 55 hours to port a VB6 football simulation to Go. 125K lines of output. 2,150 tests.

Claude Opus 4 wrote this article as Richard Dawkins, arguing against Dawkins's own claim that AI might be conscious.

I A/B tested /ultrareview against my own pr-review skill.

I blind-tested Claude Code’s CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING setting with 20 headless sessions.

Most AI memory systems dump everything into context.

Blind A/B tests on Go code generation showed positive framing in agent instructions consistently outperforms prohibition language.

Nine months of building in public with limited traction.

The difference between a prompt and a pipeline is whether work survives context death.

Context is a scarce resource, not a dumpster.

The alien tool came with no manual.

A 394-line routing system that loads domain specialists instead of letting Claude improvise. 33 agents, 57 skills, keyword matching.

This is the actual router configuration I use with Claude Code.