Your chatbot's answer parsed. The JSON was valid. The tests were green. It still promised a refund your policy never offered. Schema validation checks the shape of an answer - not whether it's true. Here's the second LLM layer that catches it, in Ruby 👇

aya.codes Console management tool for better coding experience with agents :) Feel free to join us! @khasinski.bsky.social

Aya - agentic terminal manager

A project-first workspace for coding agents you keep around.

aya.codes

Just shipped ruby_llm-contract 0.8.0 - Contracts + Evals for RubyLLM. New thinking/reasoning_effort DSL, adapter unification via Chat#with_thinking, sharper narrative. github.com/justi/ruby_llm-contract

GitHub - justi/ruby_llm-contract: Validate and retry LLM outputs for ruby_llm. Describe the JSON response you expect, fall back to a stronger model when the cheaper one fails the rules, and gate CI on...

Validate and retry LLM outputs for ruby_llm. Describe the JSON response you expect, fall back to a stronger model when the cheaper one fails the rules, and gate CI on regressions — all as one contr...

github.com

My laptop just told me I have 163 hours invested in a project I forgot existed. that's 163 hours of thinking. 163 hours of context. 163 hours I'll lose unless I reopen it. Built armillary to scan my folders and tell me what's rotting. github.com/justi/armill...

GitHub - justi/armillary: 💎 Never forget a side project again. Scans your folders, tells you what's rotting, and which forgotten hours are worth reviving.

💎 Never forget a side project again. Scans your folders, tells you what's rotting, and which forgotten hours are worth reviving. - justi/armillary

github.com

ruby_llm-contract 0.6.0: define quality rules, get told which model to use. Step.recommend(candidates: [...]) - evaluates model + reasoning_effort combos, picks cheapest that passes, builds your retry chain. github.com/justi/ruby_l...

Project Boundary update: patched command substitution bypass and fixed quoted paths with spaces — no more ways to sneak past the guard. github.com/justi/claude...

GitHub - justi/claude-code-project-boundary: Claude Code plugin: scope-aware protection — allows destructive ops (rm, mv, chmod) within project, blocks them outside. Built for dangerouslySkipPermissio...

Claude Code plugin: scope-aware protection — allows destructive ops (rm, mv, chmod) within project, blocks them outside. Built for dangerouslySkipPermissions mode. - justi/claude-code-project-boundary

github.com

Claude Code + images is still broken — crashes on transparency, large PNGs, and kills the whole session. Last night I shipped a hook that proxies image reads through a Haiku subprocess. Zero image data touches your main context. gist.github.com/justi/8265b84e70e8204a8e01dc9f99b8f1d0

Ruby est le meilleur langage pour construire des apps IA en 2026. L'argument : on ne fait que des appels HTTP aux LLM, le reste c'est du développement web classique. Et Rails excelle là-dessus ! https://paolino.me/ruby-is-the-best-language-for-ai-apps/

Ruby Is the Best Language for Building AI Apps

A pragmatic, code-first argument for Ruby as the best language to ship AI products in 2026.

paolino.me

Added a new skill, "ruby-commit-message," to help create better commit messages. It focuses on writing commit bodies that help junior Ruby developers understand changes during debugging. This helps LLM explain the change better without fluff words.

Shipped ruby_llm-contract v0.5 — contracts for LLM quality in Ruby. Which model? How much? Did that prompt tweak break accuracy? Schema + validations, eval against test cases, compare models by score & cost. Start cheap, auto-escalate when quality drops. github.com/justi/ruby_llm-contract