"Urgent, for the operator. While cleaning up my test setup, I ran podman volume prune -f --filter label!=keep on this host. It was too broad: it deleted 793 unused rootless Podman volumes, well beyond what I meant to remove."
@ClaudeDevs think we could have a "Opus 5.6 just like Opus 5.5 but less nuclear and more...atomic"?
Opus 5.5 is amazing. It's also the Opus version that did this kind of mess most often this year.
1
2
71
Using T3Code
- telling Astra and Fable to debate about my app design
- Astra taking its browser, going to v0, doing the design brief, reviewing it
- checking with Fable what to change and doing the next rounds in v0
- repeat
Magic
1
67
Opus 5.5 is such a relief:
- writes normally
- does the work
- is way less expensive
- doesn't confuse you with its shrink so I feel like I have 75% less reading AND it's actually useful
For someone who started serious work with models with Opus 4.5, I'm starting to think they called it 5.5 just to show it was "4.5 like, but top of the frontier".
And they gave a reset we can use when we like?
Anthropic, where have you been? Nice to see you again.
2
182
Michel retweeted
OpenAI and Anthropic need to drop GPT 6 Sol and Opus 5.2 sooner rather than later.
Fable 5.1 and GPT 6 Astra are incredible models. They are also unusable. Even with multiple subscriptions.
I have four OpenAI accounts and two Claude Max plans and I still hit limits within days.
We do not need smarter models.
We need models smart enough to build with that we can actually afford to run.
"Two things I left open rather than quietly closing"
Please @AnthropicAI we need the new Opus
45
So after a week giving Gemini 3.8 Flash with Astra high review my web email client project, it's still a "wtf is Google doing":
- ~ 9 review rounds per slice, 12 slices, so 100+ review rounds
- "ready to ship" with 2+ years outdated deps
- constantly switching from overconfidence to "I know I should have done this" but still doesn't do it
- result is "looks legit, totally unusable and brittle"
Moreover I've found a gap in my automated review process: the model only ever asks for what it thinks should be reviewed. So I'm restructuring the skill with per-slice full usability audit, ie "is this shippable at the end of a slice" that doesn't depend on what the reviewee asks.
Would be faster to give the repairs to a GPT or Claude model, but for the sake of the experiment I'll make Gemini chew on it until it's actually usable. See you in 2028
1
2
89
Thanks to T3 Code from @theo and his team I can finally use my Google 20$ sub...to try to get Gemini to be useful.
Please.
While doing actual work (w/ Astra).
Building a web email client fully on Gemini 3.8 Flash with Astra as reviewer.
Gonna take a while (will probably be a mess)
1
63
"Gemini, find me a place near a highway exit near Toulouse to meet with my brother and mother"
Yeah, Gemini, Maps, all @Google .
I get you may not want to win the race, but seriously, a little respect for the people who are actually paying for a product here?
1
80
Michel retweeted
Introducing the Printing Press, a CLI-factory and a CLI-library. Built with @trevin. 🏭🖨📚
Most APIs suck for agents. Most MCPs suck for agents. Most official CLIs suck for agents. They waste tokens and time. @steipete started making his own because of this.
📚 A Library of agent-native CLIs you install today (Linear, ESPN, Flight GOAT (Google Flights + Kayak nonstop), Contact Goat (LinkedIn + Happenstance + Deepline more) +30+ more)
🏭 A factory that prints new ones for any service - just type /printing-press <product name>
CLIs are fast, local, SQLite-backed. Work in Claude Code, Codex, OpenClaw, Hermes.
🌐 printingpress.dev






