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
Use the Primitive, Not the Patch: Closing a CodeQL Backlog by Class
When CodeQL keeps flagging js/file-system-race, the fix is one canonical primitive routed through one shared module. Eleven TOCTOU sites closed.
ReadOne Folder Was Behind Every Critical Dependabot Alert
When the only two manifests in your Dependabot queue belong to a folder your plugin never ships, the fix is to delete the folder, not patch it.
ReadTailscale Federated Trust: Scope to Auth Keys, Not All
Tailscale's federated trust lets a GitHub Actions OIDC subject do anything on your tailnet by default. A deploy only needs to create an auth key.
ReadPin the Installer and Add a Renewer
A CI tool that floats is a gate whose behaviour changes without a commit. Pin the installer, then add a renewer so the pin is refreshed by review.
ReadMove Deterministic Math Out of the LLM Prompt
Move date windows, site lists, and per-site totals out of the LLM prompt and into a Python file the wrapper reads before the prompt is built.
ReadRefuse to Substitute a Weaker Token in Scheduled CI
A scheduled CI workflow that falls back to a weaker token opens PRs that re-trigger no required checks. PR #1563 closes that fallback in update-npm-stats.yml.
ReadGit Plumbing for an Unattended Cron on a Shared Checkout
When an unattended cron has to commit to a shared, often-dirty checkout, git plumbing beats porcelain: hash-object + commit-tree + push survives concurrent sessions, behind-branch state, and push races.
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
August 2026: A Control That Fires on Every Case and Changes No Outcome Is a Log Line
A control that fires on every case and changes no outcome is a log line, not a control. August 2026 retrospective: 31 of 31 posts flagged for inflation with the distribution unmoved, a learned cap rule that caught zero, and a rubric six weeks stale that still names a retired rule.
July 2026July 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.
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.