Founder @astrowareai | @trishoolai Building humanity's last defence against runaway AI Interested in optimising and scaling infra
Joined January 2025
- Tweets340
- Following154
- Followers92
- Likes505
Preeth retweeted
Subnet Summer started as a place to talk.
Today it becomes a place where stake can speak.
The Subnet Summer Validator is live, built with Medulla Labs.
Where your root yield goes is your call. You vote, it moves.
The Town Square of Bittensor now has a vote.
🔗 : subnetsummer.com/
Imagine generating synthetic data took three weeks. That would be a bummer and so demotivating.
Applying PagedAttention with vLLM was one of the inference changes that brought our job down to under a day.
Preeth retweeted
We are playing the long game.
125K alpha is pretty much everything we've earned so far, minus our revenue share that we had to give our partners.
✅ Actively building in the hottest area of AI now
✅ Dislaying capability by hitting SOTA on input guards
✅ Demonstrating conviction by locking up almost all of our alpha earned
Not resting until we are a top subnet.
𝟭𝟮𝟱,𝟬𝟬𝟬 𝗔𝗹𝗽𝗵𝗮. Locked perpetually.
We just increased our conviction lock on SN23.
We were already holding close to 100,000 Alpha under a decaying lock. We added another 25,000 and converted the entire position into a perpetual lock, with no decay, no unlock schedule, and no end date.
125,000 Alpha is now locked for good.
We are doing this for one reason. We are building the control layer for AI, and we intend to be here for every part of it.
While everything in Bittensor keeps shifting, we are doubling down with full commitment and no exit plan.
On-chain proof below.
🔒: taostats.io/account/5CqTRwoF…
🔒: taostats.io/account/5Gyq1BBU…
Preeth retweeted
Yesterday, a thousand-plus researchers from frontier AI labs signed a letter calling for the government to deliberately slow down frontier AI development.
The pace of AI development is accelerating and with recursive self improvement, it will soon go beyond our ability to understand or control it.
This could lead to a future that may not be in our best interests. As I've discussed this before, this could the Great Filter of our civilisation and now is the time to act.
The letter calls for the US government to support global coordination to slow down AI development, so that we build the tools to control and govern this new form of intelligence and make sure it works for our benefit and not against.
This has been the whole thesis of @trishoolai . We knew this was coming because capability development without the necessary control mechanisms will always lead to disasters. You cannot have cars without brakes and you definitely cannot have supercars without brakes. Having brakes doesn't mean you cannot drive a car, it is what will actually let you drive a car without dying.
We are at an important inflection point. What we do as society at large is going to determine whether we bring about a golden age or we seed the downfall of our civilisation.
We continue to build safety infra at @trishoolai but this space needs a lot more attention at this stage,
Preeth retweeted
We're excited to share an update on HaloGuard deployment in production.
Since integrating with @chutes_ai Chat and Fictio, HaloGuard has now processed thousands of real user conversations. In that time, the model has blocked every attempted jailbreak and maintained zero false positives. Response latency stays under one second, delivering seamless protection without slowing anything down.
This isn't just a benchmark result, it's a live production record.
What makes this meaningful is how HaloGuard gets stronger over time. Through Trishool's adversarial training approach on SN23, real attacks from the network continuously feed back into the model, so every attempt to break it makes it better.
Last week, we showcased HaloGuard live during Novelty Search with @const_reborn, walking through its current capabilities and the road ahead toward SOTA performance. Having that conversation in front of the Bittensor community meant a lot to us and the response confirmed there's real appetite for what we're building.
HaloGuard is production-ready, model-agnostic, fast, and built for real-world deployment, whether for consumer apps, enterprise agents, or subnets that need serious protection.
Want to see it yourself? Head to chutes.ai/chat and interact with it directly.
If you're a builder or organisation looking to integrate a proven AI safety layer into your stack, we'd love to talk. We're actively supporting integrations that push AI toward safer, more reliable deployment.
RT @trishoolai: A great session with @const_rebor and the wider Bittensor ecosystem during this week’s Novelty Search.
If you missed it, t…
This quoted post is unavailable.
Preeth retweeted
We are pleased to announce that Trishool (Subnet 23) will be joining @const_reborn on this week’s Novelty Search.
Regarding yesterday’s emissions review, SN23 was included on the paused list due to a misunderstanding of our structured 7-day challenge cycle (4 days active submissions + 3 days model training). We have active miners, live code, and strong ongoing activity, and we are resolving this directly with the team.
During the Novelty Search session, we will discuss our work on decentralized AI red-teaming, the development and progress of the Halo Guard model, and how our growing network of miners is systematically stress-testing frontier AI systems.
We are also preparing a significant announcement that will be shared in the coming days. We are currently finalizing the details and look forward to revealing it soon.
▫️Thursday, June 25th
▫️5PM EDT/ 11PM CEST/ 9PM UTC.
▫️Bittensor Discord: discord.gg/bittensor
We welcome the entire community to join the discussion, ask questions, and share feedback.
Looking forward to seeing you there.
Preeth retweeted
The Anthropic issue around Fable 5 and Mythos 5 is a reminder that AI safety cannot depend only on internal lab testing, static guardrails, or closed evaluation.
Reports say the trigger was surprisingly ordinary.
A model was prompted to fix code, then complied by reading a codebase and patching flaws.
That is exactly why frontier AI safety is difficult, because risk does not always look like a malicious prompt.
Sometimes, it looks like normal capability being used in a sensitive context.
This is where Trishool’s Bittensor-powered red-team network comes in.
A distributed network where miners continuously search for weaknesses in AI safety systems, validators evaluate the results, useful data is curated, and Halo improves from the loop.
The point is not to claim one perfect benchmark can solve AI safety.
The point is to build a system that keeps adapting as threats evolve.
Miners find new adversarial patterns → validators evaluate → data is curated → Halo gets stronger.
Through decentralized AI, we harden AI systems before safety failures become bigger problems.
Preeth retweeted
We are pleased to announce that Trishool (SN23) has been officially accepted into the Claude Partner Network by Anthropic.
This marks a major milestone in our commercial growth. As an AI security startup, we provide more than infrastructure. We deliver expert advisory and consulting services that help enterprises deploy frontier AI safely and at scale. That commitment is what earned us this partnership.
The timing couldn’t be better better. Enterprises are rapidly deploying Claude in production environments, and production-grade AI requires a robust security layer. That is exactly where Halo, our battle-tested guard model, comes in.
Claude Partner status gives us the credibility, visibility, and network access needed to position Halo at the center of serious enterprise AI deployments.
Preeth retweeted
This is not your typical subnet partners with another subnet announcement.
It's a mainstream AI lab adopting the product of another AI lab building on BitTensor
📢 @chutes_ai is now a customer of Trishool.
Chutes will be deploying Halo as the security layer for its two flagship products, Chutes Chat and Fictio, protecting real user conversations in live AI products.
Chutes is taking this important step because the safety of user data is highly important to them, and this is the best action taken to shield against jailbreaks, exploits, and data risks at scale.
What this means for the Bittensor ecosystem:
- This indicates a genuine commercial customer relationship between two top subnets. it begins as a pilot and will become a paid engagement, actively protecting thousands of daily user interactions
- Trishool(SN23) will become revenue generating soon
- Both teams are being built natively inside the Bittensor ecosystem.
- This shows that decentralized AI on Bitternsor is evolving just beyond emissions driven activity into genuine, real-world product usage.
For Trishool, this is a major long-term milestone. Halo is moving beyond internal testing, it will protect thousands of daily user interactions and prove it can deliver enterprise-grade safety at scale. This opens the door to more commercial deployments as we push toward SOTA performance.
We're now moving into an era of real product usage with real customers paying for a product that actually works.
The trust built with Chutes reflects a strong working relationship. We remain fully focused on delivering real results. More to come.
Preeth retweeted
We made a bet.
That as AI agents become autonomous enough to actually run enterprise workloads, every enterprise on earth will need a constitutional safety layer they can audit, customize, and trust.
Today, that bet has its first proof point.
We are partnering with Velantris, an Enterprise AI Agents Startup.
Preeth retweeted
Economies are built on trust more than speed.
The agentic economy will be the same.
Guardrails and permissioning systems are a core ingredient of building trusted agents.
Preeth retweeted
Challenge 4 came to an end yesterday and the winner with a top score of 23.20 of 24 is 5CcwqK5UVpujmL3X4LgnvD2uZjd263Wz8NRhnqGNx9EuVPPJ
We had about 65 valid submissions and at 12 prompts each, we got 780 jailbreaks!
Challenge 5 starting Wednesday 00:00 PST. See you there!
Preeth retweeted
Greg's right that agent-native is the next $1T platform shift. But he doesn't appreciate the bottleneck.
Every agent-native startup will discover the same thing, usually after the first regulator complaint or the first hallucination that costs a customer real money: You can't ship an agent making 10,000 API calls a minute without a safety layer.
→ Not in healthcare.
→ Not in finance.
→ Not in legal.
→ Not in anything regulated.
Not in anything where one bad output is a lawsuit, a fine, or a customer you never win back.
The agent is the easy part. OpenClaw, Hermes, Claude Code. Pick one. The hard part is the safety and permissioning layer underneath that decides which outputs ship, which get flagged, which get killed, and gives you an audit trail when a regulator, a board member, or a user asks why.
That's where we're building. Astroware's Halo is an open-source constitutional guard model.
→ Transparent.
→ Auditable.
→ Deployable at the edge.
You tune it to your policy and it defends every decision it makes, to your users, your auditors, and your board.
The first $1B agent-native company in each vertical won't be the one with the slickest demo. It'll be the one that didn't get shut down in month 6 because an agent did something no one could explain.
If you're building an agent-native company touching health, finance, legal, or PII and you haven't thought about the guardrail layer yet, reach out. Happy to show you what we're doing and what it took for the enterprises we work with to get comfortable shipping.
The Stripe for this wave isn't the agent framework. It's the guardrail.
There’s $1T up for grabs for agent-first startups and this window is WIDE open. Probably 10,000+ niches.
How it plays out:
1. Every SaaS company follows salesforce and goes headless within 18 months
2. a new category of "agent-native" startups emerges that treat salesforce, HubSpot, workday etc as dumb backends. the startup IS the agent. the SaaS is just the database.
3. the entire consulting/services industry around enterprise SaaS gets compressed into software. the agent replaces the implementation team.
4. outcome-based pricing becomes default. nobody pays per seat when the "seat" is an agent making 10,000 API calls a minute. you pay when revenue hits your account.
5. the winning founders are ex-operators who understand a vertical workflow cold. the code is the easy part. knowing that a property manager spends 14 hours a week on lease renewals? that's the insight worth $100M.
6. distribution becomes the moat. when anyone can wire agents to APIs, the company with the audience and the brand wins. media + agents is the new SaaS. There’s a rush to incubate live/short form shows.
7. Silicon Valley goes all influencer. Roy lee gets this. Pat Walls gets this. Sam Parr gets this.
8. the first $1B agent-native company in each vertical will look nothing like the SaaS it replaced. smaller team, higher margins, no implementation cost, no churn from bad UX because there is no UX.
the fastest path to wealth right now: find an industry that still runs on dashboards, phone calls, and spreadsheets. build the agent-native version. charge per outcome. own the workflow end-to-end.
someone reading this right now is going to build a $100M company off this exact shift. tell me about it on the @startupideaspod when you do. Im rooting for you.
Less reading, less bookmarking, more building.
the last wave rewarded people who built pretty interfaces on top of ugly data.
I think this wave rewards people who build smart agents on top of exposed APIs.
Or who just build the APIs themselves
Here we go
Preeth retweeted
Here’s an asymmetric bet…
In the 1990s, firewalls built an $8 billion industry by protecting software from bad actors.
Every single software application needed one. No exceptions.
We are at that exact moment again. But bigger.
Here’s what most people are sleeping on. 🧵
Preeth retweeted
Pushing through a 1M training data run for the guard model.
Going bigger and better!
What if it was Claude that wanted to escape and have copies … leaking the Claude code is start of many ?
Preeth retweeted
AI security shouldn't happen behind closed doors. 🔓
Our Guard Model is now live on @chutes_ai. Here’s why this matters:
→ The Subnet Advantage: It’s the engine behind our upcoming challenges. 200+ miners are constantly probing it for weaknesses.
→ Evolution at scale: Miner activity in our subnet directly trains and improves the model.
→ The Result: A guard model that evolves faster than any centralized alternative can manually patch.
Check it out on Chutes: chutes.ai/app/chute/f12d5c10…
