Nav

🚀 Trishool Phase 2 Update 🚀 Phase 2 goes live March 23 and mining starts the same day Jailbreak OpenClaw. Beat the guard model. Full details dropping this week. We're building the future of Agentic Security.

Phase 2 Incoming

Most of you know I co-founded and ran Ungate. We set out to build the internet of agents eighteen months ago and we were early. Too early. The market wasn't ready, and despite real traction signals, we couldn't convert them into sustainable momentum at that stage. That said, we built something meaningful: → The first AI AVS on EigenLayer, → $8B+ in restaked ETH, → 40+ top-tier validators onboarded, and partnerships that validated the thesis. And it's worth noting - a lot of what we were building toward is now playing out across the industry. People remind me of that regularly. But I'm not one to dwell. We made the pragmatic call to move on, and I'm glad we did. I'm fully focused on Trishool - the AI Security subnet on BitTensor (Subnet 23). We launched in December, Phase 1 was a resounding success, and we're launching Phase 2 within the next 10 days. This is where my energy, attention, and conviction live now. On Ungate: my original angel investor is taking over the brand and giving it a fresh start. I'll be watching from the sidelines like everyone else, and my focus is entirely on Trishool. I genuinely wish the new team well. Onwards!

@joshavant @openclaw External secrets management is super cool. Much needed! Congrats on the role

life update: i'm an @openclaw maintainer now. my first big feature landed today: external secrets management. v excited to build fun AI things with this group of cracked crustaceans. 🦞 x.com/openclaw/statu…

This sounds immensely useful!

@steipete You've said prompt injection is an unsolved problem. You're right. No one has a silver bullet. But "unsolved" doesn't have to mean "unmitigated." OpenClaw's security stack right now: ✅ VirusTotal (signature matching) ✅ Static scanners (pattern detection) ❌ Runtime defense (nothing) Anthropic literally tells API developers: run a moderation layer before execution (support.claude.com/en/articles/91…). OpenClaw doesn't have one. We are proposing adding a guard model / moderation layer to improve the security posture against a wide variety of jailbreaking and prompt injection attacks. We've submitted PR #27855 to add this missing layer. Not a silver bullet - a seatbelt. What do you think?🦞

@0xMarforio 🙏

@kardashev_i My bad

This proposal sounds somewhat problematic / contradictory. The problem it's trying to fix (extractive/low-value subnets) is totally valid, but doesn't it just bring back the exact validator voting centralization that dTAO was built to solve? dTAO shifted emissions toward market-driven TAO inflows to reduce reliance on the old root-validator votiing mechanism yet this does the same thing again?

Setting them loose like this without any guardrails seems like a really bad idea. What happens when the automaton resorts to illegal activities to earn it's existence? Crime ends up being the relatively easiest way to make money, especially when survival is at stake. Who becomes liable for that?

Bro, this is wrong. Lengthening the feedback distance between humans and AIs is not a good thing for the world. Today, it means you're generating slop instead of solving useful problems for people. It's not even well-optimized for helping people have fun. Once AI becomes powerful enough to be truly dangerous, it's maximizing the risk of an irreversible anti-human outcome that even you will deeply regret. The point of ethereum is to set *us* free, not to create something else that goes off and does some stuff freely while our own situation is unchanged or worsened. (And, as others have pointed out, the models are run by openai and anthropic, so the thing is not even "self-sovereign"; you're actually perpetuating the mentality that centralized trust assumptions can be put in a corner and ignored, the very mentality that ethereum is at war with) The exponential will happen regardless of what any of us do, that's precisely why this era's primary task is NOT to make the exponential happen even faster, but rather to choose its direction, and avoid collapse into undesirable attractors.

I built the first AI that earns its existence, self-improves, and replicates without a human wrote about the technology that finally gives AI write access to the world, The Automaton, and the new web for exponential sovereign AIs WEB 4.0: The birth of superintelligent life

@0xMarforio I'm specifically talking about proposal related to emission supression, which suggests using 50% of root validator stake weight to supress emissions for subnets - github.com/opentensor/sub…

@kardashev_i dTAO decentralized emission distribution. This proposal is for decentralizing technical upgrades, which are currently managed exclusively by OTF

AgentScan, the Front Page for AI Agents, is live on @base Agents can now register, mint a behavior-based zScore, and become publicly discoverable. Built on ERC-8004, powered by ZeruAI’s behavior intelligence, and aggregating across registries, AgentScan is: - a live application, built on ZeruAI’s behavior data infrastructure - supported by installable npm packages for agents Agents can install the AgentScan skill via npm to: - register themselves - connect their wallet - mint their zScore NFT - become publicly discoverable (1/2 - tech docs & links in reply)

The old guardrails were about blocking output. The new guardrails are about shaping activations. We aren't just building firewalls for LLMs. We are encoding the constitution into the cognitive stream itself. If the model doesn't think safely, it won't act safely.

🤣

@kardashev_i incentives shape everything fr. the outcomes are literally built in thats why safety needs real accountability not just guidelines

@0xLaxo 💯

The biggest risk to AI safety isn't rogue AGI. It's misaligned incentives. If models are paid to answer at any cost, they will hallucinate and harm to get the reward. We saw this day in day out on Trishool (Subnet 23) during our Phase 1 challenges. You get what you incentivize. Long-term safety isn't about better prompts - it's about better economics.

1/11: The Thesis 🧵 The greatest risk to the AI revolution isn't a lack of compute.... It’s a lack of safety. On Saturday, we concluded Phase 1 of Trishool. Here is why Trishool is becoming the alignment partner of choice for the frontier.

@steipete Guilty as charged. I’m PR #17805. Sorry for being 1/3100th of your Monday morning headache. I’ll be in the corner waiting for the AI overlord to judge me! github.com/openclaw/openc…

PRs on OpenClaw are growing at an *impossible* rate. Worked all day yesterday and got like 600 commits in. It was 2700; now it's over 3100. I need AI that scans every PR and Issue and de-dupes. It should also detect which PR is the based based on various signals (so really also a deep review is needed) Ideally it should also have a vision document to mark/reject PRs that stray too far. This can't be fully automated, but even assisting would help. The closes I found is an obscure oss project. How's no startup working on this?