Insights
Short, practical posts on how we scope, deliver, and operate AI work for small teams — plus development updates from the AI LangRenSha arena.
The long-form engineering story behind our in-house product, published as concrete proof of what we ship.
How SOLO TECH LTD designed, shipped, and operates arena.solo.pm end-to-end: seven-layer prompt caching, a multi-vendor LLM router, SSE spectators, a long-running orchestrator on Fly.io, and a post-match reflection loop — all built around small-business cost and operability constraints.
Read the full case study →
New entries appear roughly every month. Each post is written by the founder and reflects work actually delivered in the period.
How we took a broad "we want to use AI more" brief and narrowed it into a single measurable outcome — first-reply time on customer emails — before any tool was chosen.
Read the full post →
Notes on changes to the agent memory window, cost per match, and the new spectator replay view. What improved, what regressed, and what we are watching next.
Read the full post →
The smallest workflow worth building first: structured intake form → prompt library → draft deliverable. What it replaces, what it costs, and where it still needs a human in the loop.
Read the full post →
What a clean handover package looks like — credentials, runbooks, monitoring — so the client's team actually owns the workflow after we leave.
Read the full post →
What we learned from opening the arena to the public for the first time: load patterns, unexpected agent behaviours, and the small UI changes that made the match easier to follow.
Read the full post →
Enterprise transformation programmes assume budget, staffing, and runway that micro-businesses do not have. A counter-argument, with a more realistic alternative.
Read the full post →
No newsletter provider yet. For now, the simplest way to stay in touch is to reach out through the contact form and we will email new posts directly.
