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/

Now On YouTube First Post-Halving Year: What It Revealed About TAO’s Path To 2027 | Guests: - @markjeffrey / @stillcorecap - @mogmachine / @taostats - @MaxSebti / @webuildscore - @iamtheloraxnow piped.video/owyRiM1Bk3c?si=EM6S…

Red day on the chart, green day for shipping. Not financial advice. Full weekly roundup from @SubnetSummerTAO :

Quannnn ta tum tum TUMMMMMS! Coming SOON to a Bittensor near you.

Welcome to Arena. The new trading floor for the world’s next great commodity. AI. Trade AI Tokens. See what the best traders buy. Chat, make calls, make friends, make money. Fortnightly prizes for the top of the board. Connect at TAO.com/arena and get ready.

AI agent safety is becoming increasingly important as agents take on more complex tasks and interact with external tools. @trishoolai (SN23) is making progress on one of the critical parts of that challenge: ensuring AI agents can safely handle their outputs. HaloGuard's output guard has now achieved a new state of the art, recording an average F1 score of 83.63 across public benchmarks, surpassing the previous best of 82.5. What makes this particularly interesting is how the team achieved these results. A few key details: • 0.8B model, roughly a quarter of the size of the previous leading model. • Designed to make continuous output monitoring more efficient and affordable. • Developed through Phase 3 on Trishool, using 7,500+ jailbreak attempts from 125+ unique miners over 15 weeks. The approach is worth paying attention to. Miners are incentivised to find weaknesses in the guard, and those failures become valuable training signals for improving the model. This creates a continuous feedback loop where decentralised red-teaming directly contributes to stronger safety systems. It also builds on HaloGuard's earlier progress with its input guard, which had already achieved SOTA performance at a similar model size. With the output guard now catching up, both sides are coming together to provide a practical safety layer for AI agents. Efficiency matters here. Safety systems need to be capable enough to identify threats without becoming too expensive or slowing down real-world deployments. This is another example of how Bittensor's incentive-driven approach can turn distributed contributions into measurable improvements in open AI infrastructure. Models and research paper are expected on Hugging Face and arXiv soon.

NIOME is following through on one of the clearest asks from the community. After sustained requests from holders, the first-ever $GENOME to $NIOME migration is now live. The initial 5% allows eligible Base holders to convert into locked conviction positions on Subnet 55, with GENOME held in escrow and burned only after the NIOME delivery is verified. If the swap fails, the original GENOME is refunded. This is more than a straightforward token swap. Existing Genome holders are moving into NIOME as locked participants, bringing a new group of users from the Base ecosystem into Bittensor. At the same time, the structure is designed to lock capital into the subnet rather than create immediate sell pressure. Across the full migration, as much as 1,500–2,000 TAO (over $500k) could move into long-term alignment with SN55. The first 5% is open, this is the kind of move that expands the subnet while keeping holders aligned.

Coming up in 15 minutes:

NOVA (SN68) is moving from hit discovery into hit-to-lead optimization. @metanova_labs is developing a Hit-to-Lead Optimization Agent with @chutes_ai and @onepot_ai, combining an algorithm that won a NOVA Blueprint competition with chemical databases spanning billions of compounds. The goal is to help researchers explore molecular modifications, compare strategies, and prioritize candidates based on affinity, selectivity, physicochemical properties, and synthesizability. Compounds from the Onepot databases can be delivered in 5–7 business days. There's also an important confidentiality component. With Chutes integrating domain-specific models into verified Trusted Execution Environments( TEES), the collaboration aims to support inference on external infrastructure while keeping sensitive pharmaceutical research data protected and under organizational control. What's particularly interesting is the move from an algorithm developed through open competition to a tool that can support proprietary drug discovery programs. This is another step in connecting what the network produces with the iterative work required to advance candidates. The team continues building out that path.

The cost of doing business on Bittensor this year was getting steep. In fact, the average TAO price of registering a subnet in 2026 was 638 TAO, or around $159,000 at the time of each registration. • The peak was 2,633 TAO (about $796K at the time) on May 8. • New teams would either need to be well funded from Bittensor fund or get fundraised by someone like @bitstarterAI to have a chance to start their own subnet. On October 13th the stakes changed. Prior to this change, it took 8 days for a subnet price to reach the price it was prior to the current registration. This meant a flurry of registrations could keep the price high for over a week, pricing out smaller teams. • That time is now cut in half, and the prior price only takes 4 days to be reached. • Theoretically, if a subnet registration does not take place 4 days after the last price is reached, the 1 TAO floor is then reached. Now, with this change, the latest registration price was 224.64 TAO, about a third of the average TAO price this year. This makes acquiring a subnet a lot more attainable and affordable for aspiring teams. Another change that was made was the immunity period for new subnets. Immunity is the state of being safe from being deregistered from the chain. • The prior immunity period was 120 days. Now it is 60 days. • This change puts new subnet's feet to the fire and forces them to be productive projects on the network sooner rather than later. • At 60 days, a young subnet's price matters as much as what it has built. Since all 128 subnet slots are full, the lowest priced subnet on roughly a 1-day EMA, that does not have immunity, will be deregistered when someone pays the current registration price for a subnet. What are the consequences of these changes? • Subnet registration entries are now cheaper, but turnover is much faster. • Younger subnets with weak prices and very low liquidity carry even more risk than before. It's rarely been easier to acquire a subnet, but it's never been harder to stay afloat. The chain, as always, rewards productivity, true value, and acceptance from the TAO holders. The pace has quickened, and the progress continues to accelerate. What are your thoughts on the new changes? Come to Bittensor’s Town Square at the Subnet Summer Telegram group to offer your thoughts!

🚨 Subnet Summer TAO X Spaces - October 28, 2026 | Wednesday at 5PM GMT 𝗧𝗼𝗽𝗶𝗰: 𝗧𝗵𝗲 𝗙𝗶𝗿𝘀𝘁 𝗣𝗼𝘀𝘁-𝗛𝗮𝗹𝘃𝗶𝗻𝗴 𝗬𝗲𝗮𝗿: 𝗪𝗵𝗮𝘁 𝗜𝘁 𝗥𝗲𝘃𝗲𝗮𝗹𝗲𝗱 𝗔𝗯𝗼𝘂𝘁 𝗧𝗔𝗢’𝘀 𝗣𝗮𝘁𝗵 𝘁𝗼 𝟮𝟬𝟮𝟳 The halving didn’t just cut emissions. It stress-tested every part of Bittensor at once. Four voices, four angles: capital, on-chain data, building, and validating. Speakers: • Mark Jeffrey (@markjeffrey) - @stillcorecap • Mog (@mogmachine) - @taostats | @hippius_cloud | @vidaio_ | @blockmachine_io | @say_gm_ • Max Sebti (@MaxSebti) - @webuildscore (SN44) • Tribit (@iamtheloraxnow) - Medulla Labs | @subnetsummerTAO • Bryden (@SubnetSummerT) - @subnetsummerTAO Join us this Wednesday for a high-signal conversation on what the first post-halving year means for Bittensor’s path to 2027. Set your reminder and join the discussion.

Tune in

🎙️ Subnet Summer AMA × SN55 @NiomeAI is now live on YouTube We covered: 🔄 The $GENOME → $NIOME migration (starts Oct 28) 💰 Buybacks and capital alignment 🧬 Synthetic genomics for drug discovery 🚀 Where NIOME wants to be in 5 years Full replay 👇 piped.video/rJZqFFiHxdE?si=-ZJ7…

We turned the Tesla flying car scene into a playable experience with 404. Every 3D object in the scene was built with Three.js code, written by an AI coding agent pointed at the 404 game recipe. The geometry is generated in code, while textures and skies are brought in through Atlas, allowing the entire experience to run directly in the browser. With the 404 game generator, you can build anything from a vehicle or character to a city block, full environment, or playable game. All you have to do is give your agent a simple prompt, point it to the repo, and start building. Build yours here: github.com/404-Repo/404-game…