(1/6)
Your AI agent writes the code. You scroll your phone. Then you merge code you never read.
I built hyperfocus to fix that: while Claude Code works, your terminal quizzes you on the exact change it's making.
1.0 is out today. Free and open source. π§΅
(1/6)
Your AI agent writes the code. You scroll your phone. Then you merge code you never read.
I built hyperfocus to fix that: while Claude Code works, your terminal quizzes you on the exact change it's making.
1.0 is out today. Free and open source. π§΅
(4/6)
New in 1.0:
β’ first question within seconds of your prompt
β’ it learns your project (README, CLAUDE.md, commits)
β’ questions are short and point at real lines
β’ runs with a local model via Ollama, so your code never leaves your machine
(6/6)
Try it in 10 seconds, no setup:
npm install -g @ddalus/hyperfocus
hyperfocus --demo
Code (MIT): https://t.co/qzqnauw1LD
Site: https://t.co/zMf4iDtqei
Feedback and issues very welcome.
(3/6)
Four kinds of question, all about your code:
β’ why it's done this way
β’ spot the bug in these lines
β’ what does this return
β’ which file will it edit next
Miss one? Press e for a short lesson, with your own code as the example. It comes back in 1, 3 and 7 days.
(2/6)
How it works:
β’ run `hyperfocus` instead of `claude`
β’ your prompt goes in, the agent starts
β’ a question about its plan lands in seconds
β’ as it edits, you get questions on the real diff
β’ the moment it's done, you're back with a recap
Try it in 10 seconds, no setup:
npm install -g @ddalus/hyperfocus
hyperfocus --demo
Code (MIT): https://t.co/qzqnauw1LD
Site: https://t.co/zMf4iDtqei
Feedback and issues very welcome.
Your AI agent writes the code. You scroll your phone. Then you merge code you never read.
I built hyperfocus to fix that: while Claude Code works, your terminal quizzes you on the exact change it's making.
1.0 is out today. Free and open source. π§΅
Also:
β’ --stats: what you keep missing
β’ a pre-push list of the parts you don't understand yet
β’ a VS Code panel to answer next to your code
β’ works with Codex and Gemini CLI too (experimental)
A batch of questions costs ~$0.004 on your own plan.