Jack Spirou

Founder @AgentStationAI - Al Agent Computers 🤖

[Jack Spirou @jack_spirou]

@Rainmaker1973 Strawberry Moon for a Strawberry Man @iruletheworldmo

[Jack Spirou @jack_spirou]

but not very good detail... fable scores missing... funny eh?

[Jack Spirou @jack_spirou]

soo..... which is it? Mythos 5 or Fable 5?

They are not the same model, Mythos 5 is not yet available... so what are we doing here? taking the best score from both? Benchmark-maxxing hard, eh?...

https://t.co/1IJt61kyza

[Jack Spirou @jack_spirou]

aahhh more detail here https://t.co/kBkndnNWFB

[Lisan al Gaib @scaling01]

Claude Mythos 5 and Claude Fable 5 Benchmarks

[Jack Spirou @jack_spirou]

@OpenAI @OpenAIDevs not you too...

[Jack Spirou @jack_spirou]

Claude Code is literally unusable at the moment... this amount of write errors to disk is not okay... pls fix asap @claudeai @bcherny @trq212

oh - and if v2.1.159 is a fix, you need to say so, "Internal infrastructure improvements (no user-facing changes)" means nothing. You need to reprioritize a hotfix for this, plz.

https://t.co/u7nzLH9siq

[Jack Spirou @jack_spirou]

woah, not seen this before, is it new?

[Jack Spirou @jack_spirou]

First announcement signs of Claude Opus 4.8 - Embrace long-running tasks, hopefully /goal works better and opus 4.8 can set its own goals

https://t.co/MaiufERkwt

[Jack Spirou @jack_spirou]

maybe this will be a signal to companies that ai means hiring more skilled employees to move faster with ai tools

[Jack Spirou @jack_spirou]

I feel compute constrained today.... Codex too slow...

[Jack Spirou @jack_spirou]

auto mode limping across the finish line...

[Jack Spirou @jack_spirou]

Codex is a Safety Sally
Claude is a Loosey Goosey

[Jack Spirou @jack_spirou]

I feel this too: https://t.co/UEimDleLQ5

Codex is incredible. I just want remote access like cc has and some loop or option to "keep going until done"

[Jack Spirou @jack_spirou]

Planning Mode Benchmarks from @bladnman says plan, but don't use "Plan Mode" 😂

1. Claude Code + Plan Mode: hurts, Opus −15.6, Sonnet −5.0.
2. GPT 5.4 + Cursor Plan Mode: also hurts, −11.1 drop vs no planning.
3. Cursor + Plan Mode: slight benefit for Anthropic models, Opus +1.5, Sonnet +0.5.
4. Opus 4.6 & Sonnet 4.6 have similar planning results (even with planning mode off).
5. Top result: GPT 5.4 + Codex CLI at 95.0 (X-High).
6. Gemini 3.1 Pro trails at 52–57.

https://t.co/9yvRGSBKEZ

[Jack Spirou @jack_spirou]

@kelton5020 @eddyinthebush @GaelBreton /model claude-opus-4-5-20251101

[Jack Spirou @jack_spirou]

@VAdatsi ah that explains it. I'v reverted back to /model claude-opus-4-5-20251101 so i didn't try it yet

- are you seeing a good improvement on fast mode?

[Jack Spirou @jack_spirou]

fast mode: pay 2x for the same model but faster. or just /model claude-opus-4-5-20251101