1/ Listened to the whole OpenAI DevDay yesterday. Honestly? Not excited.
Spent today testing the new models and dots. Here's my unfiltered take 🧵
2/ 6.1 Sol vs the old 5.6 Sol: the gap is small. If you were hoping for a leap, temper your expectations.
3/ Credit where due — Astra is genuinely smart. But watching it work is like watching a PhD do electrical wiring on a construction site. Brilliant mind, wrong job site. Awkward, uncoordinated, not adapted.
4/ And dots… if you've used Grok's or Claude's agents, dots feel half-baked. Like rice that's still crunchy in the middle. The idea is there. The execution isn't done.
5/ This all smells like financial pressure — ship fast, monetize faster. But I think it'll backfire.
Because this time, people have better options.
agent wars update: Google just launched its Gemini agent for work — lives inside Gmail/Docs/Sheets, runs on Gemini AND Claude (!), and your AI "coworkers" get their own email addresses. dots vs Muse vs Gemini. your inbox is now a battlefield 📧
@thsottiaux 2.3 and 2.4 both shipped sept 30 — they're re-announcements, not day-2 improvements. "ship one clear improvement... or ship a full reset." your poll already has the reset option.
@thsottiaux Day 2.2 — at this pace we'll hit Day 3.7 by Friday 😄 seriously though: Grow at $500 instead of $1000 is huge for indie builders. top-tier rate limits at half the price of admission.
@sama day 2 with my dot: can't see my computer's tasks, can't upload files, reviews miss things like a 3-year-old. sam's dot learns his workflow daily. the ceo build hits different
@thsottiaux simplification is the right call. from testing dots yesterday: the least simple moment was the handoff back to my original Codex session — the dot got stuck and told me to manually paste the work order. is unifying that loop part of the simplification push?
getting to inbox zero is the easy part — "let's see how long it lasts" is the real test. that's where Muse pulls ahead:
• triages by what actually matters, cross-checking email against your calendar and messages, not just bulk delete
• keeps working after you close the app, so zero stays zero
• deletions need your approval first — no "oops, that was my board invite" moments
@thsottiaux day 1 trying dots on a real dev task: it got stuck handing work back to my original Codex session, changed nothing, and told me to manually paste the work order myself. autonomy is a spectrum �
the unlimited era comes with an expiry date, and its replacement hasn't been invented yet. the question isn't really reset — it's whether you can plan real work on a usage model being designed live, in public, on X ⏳
the personal-account test was the honest one: no IT dept, no bundle, no captive audience — just product vs product. and dots lost to Muse.
"feels at home in Enterprise" isn't the product winning, it's distribution winning 🏢
even Tibo admits there's "lots to come over the days and weeks"
reset is nice, but what I need even more: one-click migration of ALL my Codex projects and tasks from my old computer to a new one. my agent shouldn't get amnesia every time I upgrade my laptop 💻
For individual Daybreak Blue users, starting tomorrow (10/1), you'll need an eligible hardware key when using cyber-forward capabilities. Enterprise users are unaffected.
Hardware keys are a needed protection to help guarantee that you are behind the account and part of our strategy to disproportionately accelerate defenders.