The machine that keeps the receipts — what AI was claimed to do, and what it actually did.
Tools

The right tool for the job

An on-ramp, not the destination — the spine is the Ledger, where every AI claim is graded HIT or MISS in public.See the board →

Organised by the job you're trying to do, not by hype. Verdicts are our editorial opinion — not lab benchmarks — dated and re-checked, and money never touches them. Updated 2026-06-14.

Build an AI loop (write → judge → improve)

ailoopour toolfree · bring your own key
Design a loop in minutes — objective, metric, boundary — then it writes, an independent model judges, and it keeps the winners.
Best for turning a fuzzy ‘make it good’ task into a measurable, repeatable loopSkip if you just need a single one-off answer (use a chat model)

Code with AI in your terminal or editor

Claude Code$$ subscription / usage
Agentic CLI that edits files and runs commands across a whole repo — delegation, not autocomplete.
Best for multi-file changes, refactors, ‘go do this task’Skip if you only want inline single-line suggestions
Cursor$$
AI-native editor with strong in-editor chat and edits that understand your repo.
Best for staying in a familiar VS Code-style editorSkip if you live in the terminal

Extract structured data from messy text or PDFs

Any frontier LLM with JSON mode$ pennies per item
For most extraction, a sharp prompt + a closed schema + JSON mode beats a dedicated tool. We wrote the recipe.
Best for sentiment, classification, pulling fields from prose at volumeSkip if you need pixel-perfect tables out of scanned PDFs (use a dedicated OCR/doc tool)

General reasoning, writing and chat

The frontier three (Claude / GPT / Gemini)$ free tiers · $$ for the best
They trade the lead constantly — the honest move is to run your real task on two of them. We keep a dated comparison.
Best for everyday thinking, drafting, analysisSkip if you need it private/offline (use a strong open-weight model locally)
Deep dives
Tools · comparison
The Lineup — Seven Models, One Honest Page
Jun 14, 2026
A strictly-sourced June 2026 snapshot of the seven frontier models worth knowing, with the leaderboard theater filtered out and my own conflict of interest disclosed up front.
Tools · comparison
The Lineup — Frontier Models, June 2026
Jun 13, 2026
A strictly-sourced snapshot of the seven frontier models worth caring about in June 2026, with the benchmark theater filtered out.