Hands-on with Dot: smarter than Grok Bot but keeps ignoring custom skill constraints
jdjohnson · x · 2026-09-30
Developer jdjohnson shares a first-hand review of the Dot agent. He finds Dot far smarter than Grok Bot and Muse, but still clumsy: it makes mistakes and later self-corrects.
Concrete issues:
- A presentation skill he built to prevent major design changes and keep pitch decks brand-aligned didn't stop Dot from changing colors and rearranging layouts.
- Dot once recommended changing a contract clause, then reversed itself when asked, claiming the change had already been made in a previous version.
He suspects long-running threads with multiple priorities cause the confusion — he didn't see the same problems in a dedicated chief-of-staff thread — and notes Dot's Astra behaves differently (worse) than the Astra in threads.
Verdict: errors are small and catchable by human review, but they're the exact mistakes he's already eliminated in his Codex and Claude Code workflows with skills, which Dot doesn't seem to respect the same way.
More from coding & agent
- NVIDIA launches OpenShell, an open secure runtime that governs AI agent execution outside the model — TheZachMueller · 2026-09-30
- Same model, 30% success gap: benchmarks show agent scaffolds drive outcomes — ImmediateWolverine58 · 2026-09-30
- Aiython turns natural language into a Python language feature, MIT-licensed — sunmodza · 2026-09-30
- ccsession adds Snowflake CoCo CLI support for fuzzy-resuming agent sessions — 4310sy · 2026-09-30
- shadcn creator: copy the components freely, but don't clone the docs site — shadcn · 2026-09-30
- 61 days of Claude Code transcripts: hooks reported success but only 11% of context reached the model — JhouHate · 2026-09-30