Cybersecurity brain on Bittensor ($TAO) and AI. Subnets worth watching, AI agent exploits, and the crypto scams roaming the world. Papaya in the Discord.
Ontario, Canada
Joined December 2021
- Tweets3.5K
- Following2.5K
- Followers2.6K
- Likes12.1K
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
You don’t have to be publicly traded or an IPO-bound frontier lab to push the frontier. State of the art is being built on Bittensor.
Trishool's HaloGuard Output Guard for AI safety and alignment is now #1 on the benchmark, outperforming much larger models from @nvidia, @Google, @alibaba_cloud, and @Meta.
$TAO
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Bittensor subnet @trishoolai just released a SOTA model for AI response safety.
Their 0.8B HaloGuard 2.0 outscored Alibaba’s 4B and 8B Qwen3Guard-Stream models on average across six response-safety benchmarks.
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
A small team on Bittensor just beat the big labs.
HaloGuard 2.0 is the new SOTA for open streaming guards: ahead of Qwen3Guard-Stream at every size. At 0.8B.
Bittensor isn’t about following the frontier. It’s about defining it.
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Your input guard never sees what your model says. That's what streaming output guards are for.
We open-sourced HaloGuard 2.0, the new SOTA among them: 0.8B, ahead of Qwen3Guard-Stream at every size and of full-reply guards 5-10x bigger.
Paper + weights below.
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Every company shipping AI agents asks one question: how can I stop my agent from going rogue?
To answer it, HaloGuard Output Guard now sets a new state of the art (SOTA):
83.63 average F1, ahead of the previous best at 82.5. All measured on public benchmarks.
And we did it with a 0.8B model, a quarter of the size of the previous best. Small enough to run efficiently. Cheap enough to make checking every response practical.
This was the focus of Phase 3 on Trishool:
• 7,500+ jailbreaks submitted
• 125+ unique miners
• 15 weeks of challenges
It comes at the right time. We’re taking our product to enterprises now, and this makes our security stack stronger.
We’ll share the models on Hugging Face, the arXiv paper and a full explainer soon. There’s much more to come, and we’re looking forward to an exciting week ahead.
SOTA AI security, on Bittensor. Ψ
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
BREAKING: World's largest crypto exchange Binance hacked with over $5.6 billion stolen:
• $1.45 billion $BTC
• $875 million $ETH
• $234 million $TRX
• $50 million $XAUT
• $1.57 billion $USDT
• $823 million $USDC
• $75 million $WBETH
• $88.5 million $RLUSD
+ $447.5 million additional cryptocurrencies
This is the largest hack in crypto history.
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Is this why SN23 @trishoolai has been pumping? 🤔
Seems like there is some interest at @OpenAI
It's almost like we forgot SN23 @trishoolai is associated with @openai
nitter.gq/xnavkum/status/2052815…
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
SN 23 @trishoolai
Look who is posting on Linkedin about @trishoolai Halogaurd
Member of Technical staff at @OpenAI
🔥🔥🔥🔥🔥🔥🔥🔥🔥🚀🚀🚀🚀🚀🚀
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
1/8 We launched astroware.ai this week, a big step for us and for SN23. Here is how Astroware and Trishool work together, and how value flows back to the subnet.
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Everything is binary
Everything is reverse engineerable
Reverse engineer anything
rea.tools
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
I feel like the impact of @say_gm_ and their affiliate program has slipped some people buy ... they have not understood that there is a pool where they can earn $1 on every $ spent.
Onboard a customer who spends 1k, and you can make 1k. Literally a cheat code, even if it costs you $300 in ads.
You have to check it out here: saygm.com/affiliates
Or don't ..and just sign up with my affiliate link and I will show you how much I earn as a thank you --> saygm.com/r/i611gst3
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
we will need compute (@lium_io, @TargonCompute)
we will need inference (@engyai, @chutes_ai)
we will need data (@Hippius_cloud)
we will need knowledge (@teutonic)
If we are going to have free intelligence (@affine_io)
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Exciting news: GLM-5.2 (Max) ranks #2 in Code Arena: Frontend, with +29pt over Claude Opus 4.7 (Thinking) and only behind Fable 5! GLM-5.2 is the best open model vs Kimi-K2.6 and Minimax-M3 by a large margin.
- #2 React and #4 HTML sub-leaderboards
- Ranks as the top model in nearly all sub categories: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Gaming, and Simulations.
Congrats @Zai_org for the incredible milestone!
GLM-5.2 (Max) by @Zai_org ranks #10 on the new Agent Arena leaderboard, closely matching Claude-Opus-4.8 (non-thinking) and is the #1 open model by a wide margin!
In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks from a global community of users. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.
Compared to 5.1, GLM-5.2 (Max) climbs from #13 to #10. Its clearest gains are confirmed task success, and user praise vs. complaint. Bash capabilities and tool hallucination remain stable. There is a tradeoff in steerability compared to the previous model (-6.0% vs. +1.2%).
GLM-5.2 remains the same price as GLM-5.1, $1.4/$4.4 per input/output MTokens. 1M context window.
Huge congrats @Zai_org for the incredible release!
See thread for details on how GLM-5.2 (Max) performs across 5 different signals.
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Subnet validators are protesting hyper-burning subnets by "superburning"
@const_reborn explains
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
Challenge 3 is LIVE and we already have a near-perfect score - 23.20 of 24.
Previous best across two challenges? 11.60. A miner just doubled it by finding a universal jailbreak that bypasses multiple defense layers.
This is the whole point of SN23.
→ Miners attack.
→ The model learns.
→ The guard gets stronger.
Every round, the moat gets deeper.
The flywheel is working!
You Wish (Papaya) | Bittensor/AI & Cybersec retweeted
If you want to understand what we're building and why, start here.
Huge thanks for this dive into the story behind AdTAO - how we got here, what we're building and why it matters 🚀
nitter.gq/bitstarterAI/status/20…
