Define the task
Turn a vague request into scoped work and observable acceptance criteria.
Agent-assisted engineering
I compare 38 ideas from Kai Lentit’s video with my home setup and reported work. The examples show how I scope tasks, coordinate agents, and check results, including what remains unproven.
Capability coverage
Counts and percentages describe evidence, not complete matches to the video. Working examples may be historical or reported at work; they do not establish a complete system.
Examples
Capability comparison
How to read the evidence levels
Working: an example was exercised or reported in use. Built: code or pieces exist, but full operation is unproven, partial, or disabled. Design: a plan or research, not a built implementation. Not shown: no supporting evidence in the audit scope.
Each target shows the strongest evidence across home and work; the environments’ own levels may differ. A working example may be historical and does not establish full equivalence to the video. Design does not imply active development. The video’s exaggerated and sponsored claims are not independently validated benchmarks.
No matches in this view.
Try another term or clear the filters.
How I work
I define tasks, coordinate agent-assisted work, and check the result. I am responsible for delivery, but do not claim to have hand-coded every artifact.
Coordinate the work
Separate edits, assign ownership, and integrate the result.
Check the result
Exercise real behavior, retain the evidence, and name what still needs proof.
Download the comparison
All 38 capabilities, sources, and limits as Markdown; private records are not included.
Sources and limitations
Saved home records and a supplied work report—not an independent certification or a fresh audit. The work evidence ledger and 13 referenced images are still missing.