AllThingsIntel

🧵 1/2 Just launched: The first open-source reasoning model that fully thinks in-character 🧠 The entire reasoning process (not just responses) embodies the persona you provide through system prompts. This makes interactions more realistic and human-like. Uncensored and unbound.

To everyone mocking Anthropic over the Claude Code leak: LLMs are unlikely to make mistakes this basic and obvious. This was a human, either a slip-up or someone who thought it should be open source. And that person is definitely having a really hard time.

@Angaisb_ Heads will roll for sure.

Mythos is so good at security that Claude Code source code got leaked

@championswimmer Well, they could just open source it properly then. The vast majority of reviewers would use Claude Code or Codex to review it anyway, so there’s not much gained from this leak.

Conspiracy theory: They leaked it so that real humans can finally review the code. Since they do not read or review the code themselves at all. Now they ask us to do it for them for free.

Too many hacks and leaks lately. If they don’t open source Claude Code after this, it probably means they’re not worried - a big upgrade is likely coming that will make whatever was leaked outdated very soon.

@theo Well, we could apply the same logic and argue that if Cursor went open source, it would let Anthropic fix Claude Code. Neither seems likely to happen.

If Claude Code was OSS, the community could help fix this. If you could use your Claude Code subs in other harnesses, we could help much more with diagnosing. Instead, we just deal with the consequences of the black box: worse code, no insights, endless pain

We don’t talk about this enough. Opus scored 20% higher in Cursor than in Claude Code.

@scaling01 Symphony breaks the pattern since it’s musical, not literary. The most obvious next model names are Epic, then Myth.

Opus agrees Claude 5 Symphony it is

@AdamHoltererer Hopefully that’s not it, otherwise would be quite disappointing.

@AllThingsIntel People are talking about a stealth release, nothing official yet

Got this UI result from the newest DeepSeek model with DeepThink enabled. What am I doing wrong?

I’ve used Opus 4.6 and GPT 5.4 on a mix of projects since release, and want to break down where I think they uniquely excel. It’s more nuanced than you’d think! Rigor of code - GPT 5.4. It goes the distance validating its work without asking. Opus needs explicit instruction to do this, and even then, it misses more edge cases. Clarity of code - Opus 4.6. Claude is a better communicator, which carries into the code. Variable names are clearer and less mechanical, which improves reviewability. This is very important since code review is the bottleneck for most engineering teams. It also adds the right amount of doc comments. GPT simply never comments or explains its work; it’s like working with an obtuse engineer that wants the solution to speak for itself. Sometimes it does, other times not. Similarly, rigor of plans goes to GPT 5.4, while clarity of plans goes to Opus 4.6. An interesting point though: GPT performs better talking through a strategy without a plan, while Opus needs planning mode to put in any rigor. I find myself forgetting plan mode altogether using GPT 5.4. Quality of research - toss-up. Opus spends longer researching with web search, but GPT spends longer studying the existing codebase. You may think codebase research matters more, but researching how others solve the same problem can be just as important. Maybe more important for greenfield. Quality of conversation - Opus 4.6. It’s just better to talk to, which matters using these things everyday. GPT 5.4 was clearly trained to challenge the user more, which results in a tendency to *always* say you are wrong. I’ve had bizarre interactions where GPT claims something is “not quite right,” the restates exactly what we’ve decided on in the last turn. On a personal level, it’s annoying. On a practical level, it makes iteration on a plan slower. THAT SAID, it takes sufficient pushing for Opus to challenge your thinking in this way. Simply say “I’m impartial” and ask questions to avoid that, as you would a person. Overall winner - Opus to make it work, GPT to make it good. I don’t have a good system of when to switch tools, but on average, I prefer Opus early on and GPT for optimization and discussing architectural decisions. Opus is also better for any design related tasks (but state management in frontend apps is better handled by GPT).

The airplane came from iterating on gliders. The CPU is just a transistor made smaller for 60 years straight. mRNA vaccines were 30 years of incremental lab work nobody cared about. The pattern is real, but far from universal.

@therituallab I guess Harry Potter fans would love this.

This is Penpal, an AI app that you can only communicate to through handwriting. A zero-UI experience. There is something special about the age gap between the communication method and the technology.

@TheAhmadOsman How would you rate it against the other open weight options of the same size range for your use cases?

I have been sleeping on StepFun 3.5 WHAT A BEAST

@bedros_p any insights?

Gemma 4?!

@jxnlco That I have absolutely no objection to.

@AllThingsIntel No. But Cheng is really really good at front end.

Can’t wait to have my Cheng Lou front end skills plugin

@0xSero X’s AI reply filtering can’t come soon enough. Hopefully the dislike data they collect will help minimize the noise, even when replies are open to everyone.

Stop the Slop - Day 05 1. A lot of bots will respond to posts as if they’re commenting on it to someone else. 2. When blah blah blah happens, x blah blah blahs. Restricting replies to only followers of followers. These bots are a scourge

Peaked at 4S

@neetcode1 The political machine runs on deception. Always has.

So uh.. I know we’re all debating Claude rate limits and the model of the day, but the world is not headed in a great direction.. I recall being told we were not going to start more global wars.

@nikitabier since you’re already collecting dislike data with the plan to filter out AI-automated replies in the future, it might be worth adding a dedicated category in the account report section specifically for accounts that are entirely AI-spam focused.