kachar.dev

Ilko Kacharov · CTO & Co-founder, Juma Labs

Building with AI agents, in the open.

Ilko Kacharov is the agent-native CTO: he builds Juma Labs with AI agents as part of the team, and writes firsthand about what works and what breaks.

Open the indexAbout the author

Latest dispatch

Newest
A hazy chamber of dark brushed metal. Two empty steel chairs face each other across a bare table, and a single electric-violet sphere arcs through the air between them, mid-pass, trailing light.

No. 245 min read

Orders get obedience. Questions get opinions.

I don't tell the model what to do. I ask it, because an order caps the answer at what I already knew, and a question leaves room to be wrong.

Read the essay

From the index

All 24 essays
  1. 23

    The State of AI 2026 bets your next customer is an agent with a wallet

    The 244-slide State of AI Report 2026, cut to what a startup CTO can act on: who earns the money, what an answer costs, and nine bets for next year.

    · 7 min

  2. 22

    Drift by a thousand suggestions

    I asked an AI to review the same clean file ten times and applied every change. By round three it had invented a feature. By round four it was citing that feature as my spec.

    · 11 min

  3. 21

    Fifteen years of GitHub stars, one AI wave

    I went through 2,143 GitHub stars from 2011 to today. AI repos go quiet at the same rate as everything else. What changed is who reads the README.

    · 6 min

  4. 20

    Agents should draw their work. Miro MCP lets them.

    Agents now make more than anyone can read. Connect Claude Code or Codex to the Miro MCP and have them draw it instead. Five skills to copy.

    · 9 min

  5. 19

    BM25 got Orama. Jev still needs one.

    I tried to build an Orama for Jev. It matched the best hand-tuned call on one catalog and lost by 19 points on real MCP tools. The defaults are the product.

    · 7 min

  6. 18

    Picking the right tool for an LLM is a decision. Jev makes it.

    Claude's tool search ranks with BM25. On 525 real MCP tools, Jev put the right tool first 56% of the time to BM25's 32%, and it could say none of them fit.

    · 11 min

More essays

  1. 17Your AI hub can't do the 70% · 10 min
  2. 16The lights were never off · 15 min
  3. 15Write the done-condition, not the prompt · 6 min
  4. 14Loop engineering is verifier engineering · 16 min
  5. 13Prompt engineering is over. You're a context engineer now. · 10 min
  6. 12The AI coworker is a weekend project now · 4 min
  7. 11Your product is the easy part now · 3 min
  8. 10The chat UX is the product, not the model · 3 min
  9. 09Stop picking npm or shadcn for your users · 4 min
  10. 08We deleted Jira in three weeks · 3 min
  11. 07Your homepage should run the product · 3 min
  12. 06Build for the model you'll lose · 3 min
  13. 05Agent memory is mostly hoarding · 2 min
  14. 04Context is the new CMS · 2 min
  15. 03Your evals are the moat · 2 min

Every essay, grouped by month, newest first.

Browse all 24 essays