20 Apr 2026Soon
Inference
KV cache, explained without the buzzwords
Why KV cache is the single biggest lever on LLM cost and latency, and what changes when you enable it on your own models.
9 min read
05Lab
Notes from
the workbench.Notes from the workbench.
Short, technical write-ups on what I'm learning while shipping AI products — agent architectures, inference internals, build logs, and the small field notes worth keeping.
- 12 Mar 2026SoonAgents
Agent architectures I actually ship
A field guide to the three agent shapes I keep reaching for in production — task graphs, ReAct loops, and supervisor + worker pools.
11 min read - 08 Feb 2026SoonBuild log
Vibe — behind the scenes of a 6-agent app builder
How the design analyser, planner, and code agent talk to each other, and why E2B sandboxes were the unlock that made it ship.
14 min read - 15 Jan 2026SoonField notes
WhatsApp bots that actually convert
Patterns from shipping a bot that automated 80% of conversations — qualifying questions, fallback scripts, and the small UX details that change the close rate.
7 min read
Posts go live as I publish them. Want a heads-up? Drop me an email and I'll send the first edition when it ships.
Booking Q1 2026
Have a project that needs to ship fast and work in production?
Tell me what you're building. I reply within a working day with a clear scope, timeline, and price — usually a working prototype in 2 weeks.