OpenAI's DevDay dots agent called underwhelming in hands-on comparisons vs Grok Bot
alexisgallagher · x · 2026-10-05
Hands-on takes on OpenAI's DevDay desktop agent dots describe it as rushed and unfinished: slow loading, buggy empty states, some features (e.g. Decisions API) not actually shipped, a bloated merged ChatGPT+Codex app UI, and new models less reliable than GPT 5.6.
The core design complaint: dots gives a single bot all tools and context. hunvreus argues that's wrong — one assistant juggling a week's tasks gets confused; Grok Bot's approach of multiple specialized bots ("the A-Team, not Go Go Gadget") wins. alexisgallagher's testing adds:
- Dot has the best voice UI and local computer use, but is poorly integrated with the ChatGPT/Work/Codex ecosystem and can't use a remote Codex machine independently
- Dots is less effective at using its own computer than Grok Bot, with more limits — you can't work around them by installing Tailscale or the codex CLI connector
- Muse is the most polished and consumer-friendly, yet shares functional deficits vs Grok Bot — unclear if due to a stricter security model, bugs, or model weakness
More from coding & agent
- Full Fallout Browser Game Built With Claude Opus 5.5: Zero Textures, Zero Sound Files, All Code-Generated — chrisfirst · 2026-10-05
- Eric Buess's screenless setup: one voice hub orchestrating every frontier model — EricBuess · 2026-10-05
- He turned an old laptop and Claude-written scripts into a 24/7 video line to his mom — spiritdouble · 2026-10-05
- Opus 5.5 puts Minecraft inside Slime Rancher — and it actually works — Angaisb_ · 2026-10-05
- GitHarness: agent memory as a Git repo cuts tokens 73.6% in one coding setup — rohanpaul_ai · 2026-10-05
- Claude Code directs a 3-minute talking-dog short film end-to-end in ComfyUI — Cheap_Credit_3957 · 2026-10-05