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 wall of dim terminal windows, with a single violet cursor glowing on the monitor in front

No. 266 min read

How to Burn Tokens Without Burning Out

Agents run at full speed, and so did I. This is how I slowed my sessions, my notifications and my nights down without giving up the results.

Read the essay

From the index

All 26 essays
  1. 25

    Compound engineering is our factory floor

    A ten-person team's agents run on one plugin: humans own the plan and the review, agents the rest. In one October week we merged 469 PRs.

    · 9 min

  2. 24

    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.

    · 5 min

  3. 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

  4. 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

  5. 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

  6. 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

More essays

  1. 19BM25 got Orama. Jev still needs one. · 7 min
  2. 18Picking the right tool for an LLM is a decision. Jev makes it. · 11 min
  3. 17Your AI hub can't do the 70% · 10 min
  4. 16The lights were never off · 15 min
  5. 15Write the done-condition, not the prompt · 6 min
  6. 14Loop engineering is verifier engineering · 16 min
  7. 13Prompt engineering is over. You're a context engineer now. · 10 min
  8. 12The AI coworker is a weekend project now · 4 min
  9. 11Your product is the easy part now · 3 min
  10. 10The chat UX is the product, not the model · 3 min
  11. 09Stop picking npm or shadcn for your users · 4 min
  12. 08We deleted Jira in three weeks · 3 min
  13. 07Your homepage should run the product · 3 min
  14. 06Build for the model you'll lose · 3 min
  15. 05Agent memory is mostly hoarding · 2 min

Every essay, grouped by month, newest first.

Browse all 26 essays