Pierre ⚡ Dev3o

RED → Agentic startup starter kit
@useAporia → AI product thinking partner
Dev3o → Agentic Engineering Bootcamp
Sandbox-Shell → macOS Seatbelt sandbox CLI

[Pierre ⚡ Dev3o @Pierozi]

https://t.co/MxTuEav9Dy

[Pierre ⚡ Dev3o @Pierozi]

@theo @theo don’t you think Dario essay is a funny parallel?

[Pierre ⚡ Dev3o @Pierozi]

This is bad for international stability…

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[ClaudeDevs @ClaudeDevs]

Learn more about self-hosted code sandboxes with these and other providers in the Claude Managed Agents docs:

https://t.co/uBoxh3eoyk

[Pierre ⚡ Dev3o @Pierozi]

Do not do this: ( see Fable own answer )

1.Where Fable’s marginal value is: deep analysis — root-causing a perf issue, reasoning about a data model, spotting cross-file patterns. That’s exactly what the tweet delegates to Opus. You’d be paying for the strongest model and routing the hardest tokens away from it.
2.The economics don’t even work. An orchestrator reads every subagent output and reasons on every coordination turn — with reasoning on Max, that’s where your limits burn. Orchestration is high-frequency, mostly mechanical glue. Putting your most expensive config there is the worst token allocation.
3.The correct inversion: Fable holds the plan and does the reasoning-heavy phases inline; delegate the mechanical work down — codebase exploration, file edits, test runs, boilerplate — to Sonnet or Haiku subagents. The tweet’s premise (“you don’t need its intelligence for every step”) is true; its conclusion about which steps is wrong.
https://t.co/X26YIWEg42 caveat in the tweet’s favor: if your sessions are long multi-phase refactors, the orchestrator’s decomposition quality dominates outcome — a bad plan can’t be saved by smart workers. So Fable-as-orchestrator isn’t crazy. But then Opus as the “reasoning” worker is redundant; you’d pair Fable with cheap workers, not with the second-most-expensive model.

[Dan McAteer @daniel_mac8]

This is the best way to use Claude Fable in Claude Code without immediately hitting your limits.

1. Model set to Fable 5
2. Reasoning on Max
3. Instruct Claude to run a dynamic workflow where:
3a. Fable is the orchestrator
3b. Opus does the reasoning heavy phases

Fable is so overpowered that you don't need its intelligence for every step.

Let it orchestrate Opus or even Sonnet.

[Pierre ⚡ Dev3o @Pierozi]

@daniel_mac8 Actually this is wrong and you should do the way around. Have Fable review and plan the work to do and let Opus implement it

[Pierre ⚡ Dev3o @Pierozi]

@UltraLinx https://t.co/MwfmNId1cS
The first and most advanced Agentic starter kit making possible SaaS AI agent with mutli-client, async task with event based system built in the agent loop so your Agent are proactive and always know what to do.

[Pierre ⚡ Dev3o @Pierozi]

New Claude Model Grading:
- Fable has been promoted Senior Architect
- Opus is now lead developer
- Sonnet are e2e tester / researcher / documentation writer
- Haiku search and write commit message

[Pierre ⚡ Dev3o @Pierozi]

@sflorimm One complex coding session on xhigh took 45min burn all 5h session of my max plan + $180 of extra usage. So I’m careful now Opus is doing good for most of the task. But fable found more clever optimisation in a data intensive application architecture.

[Pierre ⚡ Dev3o @Pierozi]

@theo Usually Inference has less customisation require than training. But given the size of the beast nv link is probably the only way to achieved the speed of the scale they need

[Pierre ⚡ Dev3o @Pierozi]

@awscloud
New account land with 10 Lambda instead of 1000 and make it impossible to stick a Lambda to 1.
It's been months now.

I get the security cap for new account but cmon request quota is slow and this error is absurd.

[Pierre ⚡ Dev3o @Pierozi]

@elder_plinius TLDR; « synthetic system prompt » not leak

[Pierre ⚡ Dev3o @Pierozi]

@Console_buche @gdamou_dev Focus sur le happy path en premier. Le plus tôt tu a des users le mieux c’est. Et des users payant c’est mieux. Après ça dépend si c’est du b2b complexe ou pas.

[Pierre ⚡ Dev3o @Pierozi]

@Console_buche Close beta always.

[Pierre ⚡ Dev3o @Pierozi]

For the first time I’m not using a model in Max usage for daily coding.
Fable in high seems better than opus in max. Still needs few days to found right balance.

[Pierre ⚡ Dev3o @Pierozi]

Fable Feedback:
You can feel in the response that the model behave completely differently.
It will require a lot of platform prompt engineering update ( Notion will surely loves that )

[Pierre ⚡ Dev3o @Pierozi]

If true that's a big leap.
Let's stress test it with some workflows!

[Pierre ⚡ Dev3o @Pierozi]

I burn my Max plan on 1 sessions yesterday spawning 55 Fable sub agents. So I guess we will need some adaptations in Skills to steers it correctly

[Pierre ⚡ Dev3o @Pierozi]

@GrindeOptions Exit liquidity for SpaceX ipo 🤷‍♂️

[Pierre ⚡ Dev3o @Pierozi]

If true that's a big leap.
Let's stress test it with some workflows!

[Claude @claudeai]

Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.

Its capabilities exceed those of any model we’ve ever made generally available.