Current work

Challenge: Useful workflow fixes are often too small for a full custom build but too repetitive to keep doing by hand.

Solution: Ship focused utilities through Atelier Tools, improve them from real usage and QA, and use repeated demand to identify larger product opportunities.

Result: A public proving layer for Ahoy’s tool-building engine, with current attention on reliable PDF utilities and a freemium/pro path.

Challenge: Sensitive document workflows need speed and structure, but generic cloud-AI promises create risk when files, originals, signatures, and review decisions matter.

Solution: Use Foundry to organise, check, preserve, and prepare document outputs locally, while keeping human approval and claim boundaries explicit.

Result: V1 release-candidate infrastructure and Pack System v0 are accepted with guardrails; PDF Ops is demoable in private/local buyer demos, not yet production route-ready.

Challenge: Local model claims get vague fast: best for coding, fastest, most reliable, all-rounder. Without visible test records, those claims are just vibes.

Solution: Publish the current tested evidence with model cards, verdict caveats, raw evidence excerpts, and methodology notes.

Result: /local is live with current recommendation cards, evidence links, and explicit limits where the data is not broad enough to crown a winner.

Challenge: AI-assisted work breaks down when one agent invents strategy, writes copy, builds UI, and approves itself.

Solution: Split the work: research maps proof, copy defines claims, design shapes the interface, build implements, and QA checks what can be verified.

Result: Faster delivery with clearer accountability — and fewer unsupported claims reaching the public site.