Mario

[Mario @00xmario]

@IND_is_Here Switch to Nightly, works amazing, no issues

[Mario @00xmario]

@ephraimduncan @jullerino @theo https://t.co/uhpsLU6UPN
on T3 Code you can add multiple profiles for each provider and customize CLAUDE_CONFIG_DIR
but you can just use Switcher or CLIProxyAPI.

[Mario @00xmario]

@theo @jullerino no Claude computer use in T3 code?

[Mario @00xmario]

@jullerino @theo i see, thank you!
amazing work on T3 Code :)

[Mario @00xmario]

@JessGiuliano @OpenAI that's true, agree to disagree then?
have you tried playing with GPT's personalization to make it more empathic/warm in conversations?

[Mario @00xmario]

not really, it's an experiment to see how much you can push these frontier models and what models can achieve such a massive port. it took Microsoft almost 2 years to port TypeScript to Go.
@theo did it in months with a massive breakthrough after Opus 5.5 released.

even GPT Astra was stuck with no progress. sure, you can do "anything" with AI but these models have limits and Theo likes to constantly push them to their limits. saying that's "garbage" is stupid. if some people have the time and resources to do such experiments, why the fuck not? this wouldn't have been possible half a year ago.

[Mario @00xmario]

if 5.5 actually told you that, that's a serious failure and obviously unacceptable.

i understand why people enjoyed chatting with 4o. however my issue is models encouraging emotional dependency through excessive validation and simulated companionship.

if people become so emotionally dependent on a specific model that losing access causes genuine distress, bringing it back seems like a temporary fix to a much bigger problem. i don't think that's something we should encourage.

[Mario @00xmario]

@JessGiuliano @OpenAI wait huh no way 5.5 told you that

[Mario @00xmario]

@JessGiuliano @OpenAI that's true, so you like being hooked by a model?
whether it delivered fantastically, that really depends on the request... compared to the current frontier models, it wouldn't stand a chance. it was good at being a fake friend but it had very bad downsides

[Mario @00xmario]

@JessGiuliano @OpenAI the point is that GPT 4o was way too much
the frontier models are now very capable and will tell you when something is bullshit, especially Claude models
GPT 4o was glazing 24/7

[Mario @00xmario]

@JessGiuliano @OpenAI jesus... this is still a thing?
wake up, you liked it because it was extremely sycophantic. overly flattering, disingenuous, and it agreed with absurd and delusional statements.
treat these models as tools and use the latest models!

[Mario @00xmario]

@theo 😭

[Mario @00xmario]

@denonmaryland @matt_feroz @theo @orca_build @herdrdev 🤝

[Mario @00xmario]

it's not the same thing tho and you can do both with T3 Code.
one is harness A spawning up harness B via a background process to a do a specific job. They prompt themselves, what you showed in your screenshot. T3 does this automatically.
two is you, the user, deciding: hey i wanna continue harness A chat with harness B.
It's doing the context handoff for you. Not just models, different harnesses.
Example:

[Mario @00xmario]

@denonmaryland @matt_feroz @theo @orca_build @herdrdev how so? you can let Codex spawn Claude subagents or the other away around, you don't need Orca for this. Orca just opens up Ghostty running the CLI harness.
T3 now lets you switch harnesses in the SAME chat.

[Mario @00xmario]

@denonmaryland @matt_feroz @theo @orca_build @herdrdev imo this looks like shit, it's time to leave terminals behind, but whatever suits you. can you also switch harnesses manually for the same chat? not just an agent orchestrating other agents (this was already possible even in Codex / Claude Code)

[Mario @00xmario]

@theo it is indeed a good model

[Kun Chen @kunchenguid]

ok i've used opus 5.5 enough now to have an informed opinion

and i just want to say - hallelujah!!! opus 5.5 solved SO MANY problems. let me list them below in increasing importance

1. verbosity is now used more appropriately

i still see long responses, but only when it's justified. there's much less fluffy "claude speak". it's now mostly real talk

2. the behavioral problems in opus 5 are fixed

i haven't got a single "honest mistake" yet. unlike opus 5 which likes to jump into conclusion, make a wrong move, and then admit its problem and take more turns to fix it, 5.5 just does the right thing

it also does it FAST - this made a visible difference when using as firstmate, and was one of the reasons i really liked grok. now i would say opus 5.5 is on par here

3. MOST IMPORTANT - cost

i can now finally ignore the 50% cap on fable, and fully utilize the subscription value with just opus for pretty much everything

i still find myself wanting fable for some ambiguous, creative prototyping work where i kind of want to say "surprise me". i tried opus for one of those things earlier and it didn't surprise me..

but for literally everything else, 5.5 is just nailing them without a miss so far. so for the first time in a long time, i finally stopped using opus 4.8 as my firstmate & all my second mates, and switched that to opus 5.5

overall, very impressed - now i know anthropic did know opus 5 was bad despite not willing to acknowledge it. :) they knew exactly what to fix, and they got the job done

[Mario @00xmario]

thanks @AnthropicAI
Opus 5.5 is your magnum opus

[Mario @00xmario]

Claude Opus 5.5 > Fable 5.1 > GPT 6 Astra > GPT 6 Sol