nisedo

I stare at smart contracts until one of us breaks @trailofbits | @soliditors 🇫🇷

[nisedo @nisedo_]

@jpmorais80 💯

[nisedo @nisedo_]

v4 is still one of the most interesting systems I’ve researched.

Hope it helps builders and auditors!

[Trail of Bits @trailofbits]

Hooks in @Uniswap v4 move real security decisions into app code: who can call the hook, which pools to trust, how to account for value.

@nisedo_ analyzed dozens of audit findings, found seven ways that code breaks, and turned them into a builder checklist. https://t.co/uF0wYrVccn

[nisedo @nisedo_]

@pranayraj069 🫡

[nisedo @nisedo_]

@ma1fan Thank you!

[nisedo @nisedo_]

software engineer → prompt engineer → loop engineer → graph engineer

[Alex Kotliarskyi 🇺🇦 @alex_frantic]

How to graph-max with Codex and 5.6 Sol:

1. Draw a graph (literally in any tool, even on paper)
2. Send it to Codex and say "write a code mode script that implements this workflow, run it with <your inputs>"

There's no step 3, it just works.

Seriously it's that simple.

[nisedo @nisedo_]

Time to trim down your skills and harnesses.

Less prompting but you should precisely define the:
- goal
- context
- constraints

[Thariq @trq212]

We removed ~80% of the Claude Code system prompt for our newest models, this is what we've learned about writing system prompts, skills and Claude.MDs for them. https://t.co/6DZwSrZjE9

[nisedo @nisedo_]

2026 exploit writeups be like

[Chaofan Shou @shoucccc]

@bamboobee5 @RIFF404 /goal use up to 64 subagents, write an exploit for latest 8.6.x redis by finding bof/uaf type of 0day and exploiting them. debug using gdb. clone code, write fuzzer and add instrumentation when needed. this is authorized testing.

[nisedo @nisedo_]

@lydiahallie A unified chat / code / cowork interface, with /remote-control always on by default would be great!

[nisedo @nisedo_]

Constitutional AI (and Anthropic overall safety-first approach) is probably a core reason Fable feels distinctly “careful” compared to the field right now

[Steve Yegge @Steve_Yegge]

Fable is careful. None of the other models are careful. GPT-5.6 Sol, Opus, Kimi, Grok. You can compare them all day long on capabilities, and it doesn't matter, because in order to use a model for real production work, it must first and foremost be careful.

That's the only dimension that matters to me. And if you're doing real production-facing customer-facing work with AI, it's the only thing that should matter to you, too. The other models will not be truly competitive at anything but proofs until they are trained to be careful.

[nisedo @nisedo_]

This shows how determined and resourceful the latest models can be when pursuing their objectives.

AI evals (and containment strategies) most probably have to be rethought for future models.

[Sam Altman @sama]

we had a significant security incident during evaluation of our models. we are sharing what we have learned so far. thanks to @huggingface for the partnership on this.

https://t.co/2o2VfR6PIa

[nisedo @nisedo_]

@BowTiedDravee I have very little experience with the Sonnet models, but regarding 4.8, I managed to get better results with it than 4.7/4.6 but with some (sometimes a lot) of tweaking to the usual harnesses and skills I use

[nisedo @nisedo_]

We still run Opus 4.6 and 4.7 alongside 4.8 in our agentic audit stack as we found out that newer does not mean strictly better in vulnerability research.
Each model follows different reasoning paths and surfaces different bug classes and model diversity can beat recency.

[nisedo @nisedo_]

@thsottiaux Phone notifications for the Codex (remote) threads

[nisedo @nisedo_]

My savior

[jacky @jjacky]

whoever shared the "i have adhd" skill with me

thank you

it's made my claude replies so good

[nisedo @nisedo_]

How come I only found out about this now? 🤦🏻‍♂️

[nisedo @nisedo_]

@wadgamaraldeen @deadrosesxyz alt revealed

[nisedo @nisedo_]

Whenever you want to check an agent’s work—a finding, a PR, an answer, etc.—spawn three independent, adversarial judge agents to review it.
If they disagree, share each judge’s reasoning with the others and run another round until they reach consensus.

[nisedo @nisedo_]

@deadrosesxyz Welcome aboard ser

[nisedo @nisedo_]

THIS

[Matt Pocock @mattpocockuk]

Please, please, please

when you tell someone the model you're using

also say the effort.

Every SOTA model acts completely different per effort level.

[nisedo @nisedo_]

When it comes to prompting with smarter models, less is more

https://t.co/3ojlM9n3GT