First independent Dots review: "too buggy to recommend now," says Every's Dan Shipper
Every's Dan Shipper tested Dots for several days and says it's too buggy to recommend — frequent permission failures, dropped messages, in-app browser disconnections — and advises waiting a week or two. Boo, his dot, still became a habit.
What happened
Every editor Dan Shipper published "Vibe Check: OpenAI DevDay 2026" on September 30, 2026, with the first multi-day, hands-on test of OpenAI's Dots agents — a genuinely independent evaluation, days after launch.
He used a dot named "Boo," sitting in ChatGPT's redesigned sidebar, for real work over several days: orchestrating ChatGPT threads, reading email, using Slack, filtering incoming requests to triage his inbox, and once warning him that a flight reschedule conflicted with a meeting request from Slack he hadn't read.
The verdict is harsh: "the version I tested over the last few days was too buggy for me to recommend now." Specifically, he reported frequent issues with permissions and dropped messages, and his dot was often unable to connect to the in-app browser. Dots also add complexity: he and his dot kept getting confused about whether work should happen on the dot's own cloud computer, in the in-app browser of the main thread, or in a separate ChatGPT thread the dot had kicked off and was orchestrating.
His recommendation: "unless you love new things with rough edges, I'd recommend waiting a week or two to try it," predicting OpenAI will fix many of these issues quickly. He also noted that if you're already deep into Grok Bots, Muse, or Instinct, Dots probably isn't worth switching to unless you're a heavy ChatGPT or Codex user.
It wasn't all negative: he kept reaching for his dot over starting new Codex threads, "so it has definitely become a habit despite the frustration." And ChatGPT Space got the stronger review — he wrote the entire piece in Space, calling it an easy drop-in replacement for his old document editor.
Why it matters
This is the first independent, multi-day field test of the product OpenAI is betting its enterprise future on — and it confirms what the DevDay demo stumble ("Dottie's having a slow morning") hinted at: background agents live or die on reliability, not capability. For a product whose pitch is "keep working while you sleep," permission failures and dropped messages are the worst possible failure mode, because the user isn't watching to catch them.
The review also validates OpenAI's own internal prioritization problem: Shipper's biggest praise went to Space, the comparatively boring document layer, while the flagship agent still needs a week or two of fixes. Every is a pro-automation outlet (Shipper has attended every OpenAI DevDay since 2023) — a "wait" verdict from that camp is a meaningful signal, not a hater's take.
Verified Sep 30, 2026.
Sources:
- Every (Dan Shipper), "Vibe Check: OpenAI DevDay 2026" — https://every.to/vibe-check/vibe-check-openai-devday-2026