Search
Start AI Tools with Intent Solutions
Daily work journal below. Real case studies, deep-dives, and operator-grade write-ups on AI engineering, DevOps, and the messy reality of shipping software that touches LLMs, by Jeremy Longshore at Intent Solutions.
Curated long reads
Let the Model Judge. Make the Code Decide.
The model owns judgment. Deterministic code owns the gates, the commit, and the audit trail. How I rebuilt a daily writing pipeline so a hung session can no longer brick tomorrow.
Adversarial Review: The Six Lenses That Halted a Rollout
Six lenses, live state, eighteen risks. I thought the safety work was shipped. The review proved three assumptions were theater. The durable rule: the model proposes; the system owns identity, trust, and state.
The Moat Is the Trust Layer
A local-RAG demo became Intent NEXUS. The moat is not the chatbot. It is egress control, code-enforced citations, a ledger you can check, injection defense, and evals that cannot lie quietly to you.
Recent posts
Every Check Should Report What It Did Not Look At
A status job that always runs reports which checks executed and which are unproven, so a skipped GitHub Actions job cannot read as a pass.
ReadWhat a Skipped Check Is Worth in CI
A gate runner that counts SKIP as pass prints green for work that never ran. How a rig proof receipt closes that gap, and why filters must invert.
ReadA Closed Epic Is a Claim, Not a Fact
Three agents re-executed 37 closure claims on four shipped epics and refuted three. An audit briefed to refute, not confirm, finds what a review misses.
ReadThe Refusal Nobody Heard
A pipeline guard correctly refused a diverged tree. Two nights of silence showed the real gap was that nothing was wired to listen.
ReadMake the Guard Prove It Can Fail
A guard with no reachable red path is not a guard. Four gates in one day whose verdicts were decoupled from the thing they claimed to measure.
ReadSeventeen Spellings of the Same Number
A checker that governs published skill counts must discover them without hardcoding variable names. Every commit found a new legal spelling to handle.
ReadExit Zero Can Lie; Stdout Holds the Answer
A CLI can print its diagnosis to stdout and still exit 0, where a handler watching stderr never sees it. Three days of cron packets said nothing about why.
ReadSeries & curriculum
MCP for Beginners
End-to-end Model Context Protocol curriculum — solutions in Python, TypeScript, Java, Rust, C#, .NET — translated across six languages.
Agentic Design Patterns
Patterns and anti-patterns for production agent systems. Decision frameworks, prompt scaffolds, evaluation harnesses.
Tiny Recursive Models
Building small recursive systems where simple loops compose into emergent behavior. Series in progress.
Research & Curriculum
Long-form research articles, learning paths, and reading lists for the AI builder community.
Retrospectives
July 2026: Every Instrument Reported Success, and One of the Published Posts Was a 404
July 2026 retrospective: Tier 1 collapsed from 46% to 17%, the anti-inflation rule adopted July 3 fired zero times all month, the grader demoted half the catalog on a title keyword list, and a post with a full green audit trail was never published at all.
June 2026June 2026: The First Full Month Under the Repaired Rubric — Tier 1 Went From 5% to 50%, and the Last Mile Got Louder
June 2026 retrospective: the first full month running under May's repaired tier rubric. The distribution corrected from 5% to 50% Tier 1, exactly as predicted — but the structural sweep still flags every disagreement as inflation, and no human has adjudicated one. 27 posts, 1,166 commits.
Let's build something that ships to production
Intent Solutions designs, builds, and ships production AI systems: Claude, MCP, RAG, agents, and the boring infrastructure that makes them trustworthy. Tired of AI demos that never make it past the POC?
No commitment required. We respond within 24 hours.