TeksiEvery action is a thread.
The code tells you what was built. Teksi shows you the process behind it
Teksi weaves every prompt, edit, terminal command, and AI interaction into a replayable record of how software was built — turning thousands of events into evidence your reviewers can trust.
Completed in 6.5 minutes. AI contributed 2 of 17 total actions. Final overall score: 60/100.
Engineering work is woven as it happens — not remembered afterward.
The interview question
“So — how do you usually use AI when you code?”
Told after the walk. Unverifiable. Answered the way that sounds best.
Fact · session #04821
“AI-generated code made up 34% of the final submission, concentrated in boilerplate and test scaffolding.”
Derived from captured events. Every thread traceable back to the session.
The process
How Teksi works
01
Configure the assessment
Design a task and what a strong submission looks like, optionally create a starter git repo and/or give us a Test suite (will run after the final submission and results will be aggregated in the final report)
02
Candidates work, normally
They code in the provided environment. Every prompt, paste, edit, run, and undo is captured as an event — nothing to install, nothing to remember to record.
03
Reviewers get Facts, not footage
Raw events are distilled into a structured report: what was written, what models were used (with token usage), what was accepted as-is, what was reworked, red-green test cycles and more.
Same events, different bar
Scored for what you actually hire for
The six category scores below come from one session on "react-counter-challenge". They don't change — only how they're weighted does.
No summaries. No scores you can't trace back.
What a Facts report actually says
1 prompt sent to a single build agent — no multi-agent orchestration.
15 manual edits vs 2 AI-driven actions — candidate favoured direct control.
3 unique files touched with 6 saves, 11 terminal commands.
2 test runs with 1 red-green cycle before passing.
0 errors; session completed and submitted in ~6.5 minutes.
3 agents created (debug, research, review) — toolbox score: 75/100.
Building this with hiring teams, not for them
We're working with a small group of companies to shape the first version of Teksi. Tell us what you evaluate candidates on today, and we'll show you what a Facts report would look like for it.