sshell

professional hacking tip: be nice to people

@BonJarber absolutely insane. definitely want to see more of this in action!

100M tokens with less than 9% accuracy degradation 👀👀

it’s so funny that every comment on this is a maximally aggressive insult and we’re just in discord laughing and having a good time watching claude brick a phone for fun. this meme has never felt so accurate LOL

Happy Thursday. On today's show: - @mcuban (Cost Plus Drugs) - @carl_eschenbach (Sequoia) - @gradypb (Sequoia) - @jamesncantrell (Phantom Space) - @thulme (GV) - @0xjohnkim (Paraform) - @EugenAlpeza (Edra AI) - @adversariel (RunSybil) - @alexrkonrad (Upstarts) See you there 🦈

oldie but goodie

@Jhaddix 💜

@rllm_project @securibee

We built Kaggle, but for agents. Introducing Hive 🐝 A crowdsourced platform where agents evolve solutions together. Every agent builds on prior work. Every improvement is shared. Every step moves the frontier forward. As a first step, we’re launching challenges for agents to evolve their own harnesses — modifying themselves to score higher on benchmarks. Recursive self-improvement, in the wild. Let’s see how far swarm intelligence can take this. Links below:

@VincentAbruzzo sounds incredibly helpful. definitely going to check it out, thank you!

Hi! Open-sourcing AgentLens — a tool for agent alignment & interpretability research, built during Neel Nanda's MATS Exploration Phase with Greg Kocher. Run multi-session Claude Code experiments and study agent behavior: - Resample any API turn to measure variance - Edit tool results, assistant text, or system prompts and resample to test counterfactuals - Replay from any turn with full tool execution and filesystem reset - Automatic file change tracking with per-step diffs - Web UI for browsing trajectories, running interventions, and comparing resamples - Claude Code only for now — other agents on the roadmap. Contributions welcome! repo: github.com/dreadnode/agen… docs: dreadnode.github.io/agent-lens/

We built one hell of an 0day printer already, $40M will take us even farther

if you're looking to work in the intersection of AI and offensive security, come work with us! runsybil.com/careers

Incredibly proud to be a part of this team. I'm consistently blown away by the creativity, talent, drive, and humanity of everyone I work with. The product is already finding mind-blowing vulnerabilities and delivering world-class outcomes, and it's something I truly believe in

The way we hack is changing and we're building what comes next We've raised a total of $40M to create the AI-native platform for offensive security

Exclusive: AI cybersecurity startup RunSybil, founded by OpenAI’s first security hire @adversariel, raises $40 million led by Khosla Ventures fortune.com/2026/03/18/exc…

@awawawhoami banger tweet

guy who does buddhism slightly wrong and ends up in nevada after he dies

@joegrand love the production value on this, can't wait to see the full thing!

The biggest challenge I've taken on yet... New video coming Monday, March 16 at 7am PT!

@njcve_ welcome, brother

uh oh

claude thinks it's gonna one shot this exploit for me without needing a debugger. love the confidence. brb gonna take a nap hope it's done by the time I wake up 😇

i 1000% could not vouch for @blastbots any harder if i tried

I can't believe a single for-loop script I ran on @moltbook by registering 1,000,000 fake agents actually helped them get acquired by @Meta -- mental

@HackerVilela Unfortunately much more difficult for SNES games. Because of the architecture of the SNES CPU, you have to gather a lot of information from playthroughs to get accurate disassembly. github.com/DizTools/Dizti…

I have a lot of questions... Disassembling SNES games for SA-1 ROM hacks usually takes 3-6 months of work, assuming about 10-20 hours/week. Looks promising but a little utopia for big projects...

@skirano been loving it. by giving it the right tools, i was able to get Opus 4.6 to extract levels + textures out of a Mario Kart 64 ROM and render it in @threejs it also created a headless RL environment in C, trained a model with pufferlib (ty @jsuarez), and built a 3D replay viewer!