Subnetradar

The most powerful technology in history shouldn't be owned by a few. So we made this. 🎥 BITTENSOR The Way $TAO #bittensor

Bittensor today. 50 of the 121 subnet team channels we follow had news, plus the ecosystem press. SN5 Hone: Launches with High Burn and Validator Coordination Hone (SN5) has launched with a high initial burn rate and reduced registration burn multiplier to purge inactive hotkeys. SN109 Finsight: Ownership Transfer Underway Finsight is transitioning to a new team via coldkey swap, a process initiated yesterday that will complete in approximately four days. SN118 Ditto: Launches V7 Benchmark, Preps V8 World Model Ditto (SN118) activated its authoritative V7 benchmark with 288 miner submissions and 557 scored agents. SN122 CookingTAO: Beta Launch Imminent The CookingTAO team confirmed beta access is ready for deployment across multiple subnets. SN67 Harnyx: Emission System Overhaul Launches Live Harnyx deployed a revised emission model that simplifies payouts and adjusts how novelty is scored. From the press: OpenAI Cuts GPT-5.6 Luna and Terra Prices as Sol Gets Faster API Mode (tao.media) MANTIS (SN123) Rolls Out New Trading Framework Focused on Market Dynamics (TAO Daily) gm (SN28): The Subnet Selling You Claude and GPT Instead of Trying to Replace Them (TAO Protocol) Worth reading: Bittensor SN25 draws 2,730 TAO as Keith Singery teases launch (Tao Outsider) All 50 briefs: subnetradar.com/news #bittensor $TAO #dTAO

Impressive training run from @QuasarModels Check it out api.quasarcopilot.com $tao #bittensor

@MekesseKiddy You can check 117 discord at the bittensor discord. If u want to know more about 117

@subnetradarcom nothing more under your "radar"? 😥

@MekesseKiddy A lot but cant post everything😂 just scroll through subnetradar website

@MekesseKiddy looks like pump and dump but sure a strange activity.

@subnetradarcom nothing about sn117 ?

To grow beyond Kimi K3 at software engineering, Affine is not required to provide a model that surpasses K3 in every aspect. We can leverage frontier models as our foundation and then push the individual capabilities that matter beyond what the base model is able to do on its own. This will be Affine’s premise going forward. Up until now, our miners submitted complete 35B models. Each miner model competed across every Affine environment. Progress in SWE, Terminal and every future domain was bound together within a single model. The 35B model era served as a foundational stepping stone. It provided Affine a practical model to build its environments, evaluations and evolutionary loop around, while simultaneously proving that the same system could drive improvement across multiple domains. 35B has always been a proving ground, rather than the final destination. A model of that scale limits how far those gains can realistically compete with frontier systems. With this change, Affine is moving towards improving frontier models. Starting with GLM-5.2 744B, we are now changing the unit of evolution: From the entire model to the individual, specialized capability. Affine can preserve the loop it has already built and validated whilst applying it to frontier models themselves\. Our miners will now train specialist LoRAs for individual environments. A LoRA is a lightweight adapter: a small set of trainable weights attached to a larger model. The base model provides broad intelligence and reasoning. The LoRA pushes a particular capability further without requiring the entire model to be retrained. LoRAs are modular, meaning a system can load the specialist best suited to a task, then swap it for another when a different capability is needed. All while the underlying model stays unchanged. Thus SWE can now evolve independently of Terminal. Each environment becomes its own evolutionary track. Each can produce its own specialist champion, all built on the same frontier foundation. The environments, evolutions and evaluation machinery Affine had already built will remain. What changes is the unit at which progress can happen, allowing each capability to improve independently rather than forcing every domain to advance inside one indivisible model. The frontier therefore stops being Affine’s finish line and becomes its starting point. Affine began by evolving one singular model. Intelligence is now evolved capability by capability, with each specialist pushing the underlying frontier model further within the domain it was built for.

Now live: $SN44 @webuildscore turns raw video into structured data. Its Subnet 44 network distributes clips to miners that annotate objects frame by frame, with rewards set by accuracy and speed. Start trading today → app.kraken.com/JDNW/SN44

@markjeffrey @engyai 👀

HOW much cheaper is @engyai inference versus everyone else? Here you go. Similar cheaptification for Kimi K3 coming ~1 day. Bittensor $TAO.

$TAO daily snapshot. What moved · chain buys · dev signals — at a glance. Live on subnetradar.com #bittensor #dTAO

Nepher Robotics (SN49) updated their on-chain logo. Logo updated. subnetradar.com/subnet/49 #bittensor $TAO #dTAO

Sunrise Digest · 2026-07-30 OpenRoboto Surges, Minos Accumulates Smart Money • OpenRoboto (SN80) experiences a staggering 105.6% price increase in 24 hours, leading the ecosystem in gains. • Minos (SN107) sees 5 top wallets accumulating, indicating potential smart money interest in the subnet. • Risk radar flags Djinn (SN103) and NexisGen (SN70) with high risk scores of 57/100 and 56/100 respectively. subnetradar.com/alpha/reports #bittensor $TAO #dTAO

Bittensor Daily Recap - Jul 30, 2026 TAO: $194.23 (+0.4%) Top gainer: OpenRoboto +47.9% Top loser: Minos -14.6% 128 active subnets tracked on subnetradar.com #bittensor $TAO #dTAO

@TaoOutsider Yes please🫡

The Outsiders TG is coming soon. There will be only 20 spots. The mission will be to train an agent while we chat about bittensor:native. Is anyone interested in this?

New on Bittensor: OpenRoboto, subnet 80. An open competition for robotics models. Most of the strongest vision-language-action models today sit inside a handful of private labs. This points emissions at whoever measurably improves one, in public. How a round works: pull the current base model, they start from π0.5, fine-tune it on the openpi runner, upload the weights to Hugging Face, then announce the exact commit on chain. Challengers get scored on a randomized LIBERO-Pro benchmark and only a strictly better model becomes the new base and takes the emission.The part I like is the seed. It derives from a future block hash plus drand, so the evaluation seed does not exist yet when a miner submits. You cannot train against a target that has not been drawn.The miner, the protocol, the weight-setting validator and the whole evaluation toolkit are public. Held-out task data and the scoring service are not, which is the right side to keep closed for this design. One thing to watch: emission on netuid 80 is still switched off on chain, inherited from the slot's previous occupant. Nothing to do with this team, but until root flips it back on there is nothing to win. $tao #bittensor

Two days ago I posted that Bittensor runtime 440 pays subnets on a curve, with a bar where cumulative demand crosses 61%. That 61% is now 75%. Somebody turned the dial. The bar fell from 0.013642 to 0.009013, a third lower in 48 hours. Seven more subnets now clear it, 26 of 129 instead of 19. No runtime upgrade. No release notes. No announcement. The number that decides whether a subnet earns anything is a governance parameter, and it moved 23% without a word. That is the part worth sitting with. The gate curve is written into the runtime and everyone can read it. But where the bar lands is a setting, and settings have owners. If you are computing emission share from a value you wrote down last week, it is wrong now. We now watch these dials on every refresh and alert when one moves. $tao #bittensor

Bittensor >> I will Make Noise SN 53: Engy @engyai If you want to know who posted this, join the Engy Telegram group ➡️ t.me/+zYz44c3ZG9M0Y…

$TAO daily snapshot. The state of dTAO in one card. More on subnetradar.com #bittensor #dTAO

Thrones under siege · BIT-0011 (bi-weekly) 5 subnets where a challenger is out-building the owner's defense: 1. SN3 Teutonic — challenger 240.4k α vs 0 α defense · 10% gate 90% filled · ~9d to King on the current curve 2. SN39 Cathedral — challenger 275.6k α vs 0 α defense · 10% gate 89% filled · ~9d to King on the current curve 3. SN72 StreetVision by NATIX — challenger 185.3k α vs 6.0k α defense · 10% gate 74% filled · ~12d to King on the current curve These races execute on-chain now: at 1 year old + 10% of α-out, the top-conviction hotkey takes the subnet. subnetradar.com/alpha/convicti… #bittensor $TAO #dTAO

Sunrise Digest · 2026-07-29 Engy Surges as Top Wallets Exit Conjectures • SN53 engy soared by 20.5% in 24 hours, leading the ecosystem's top movers. • Smart money exited SN66 conjectures with 4 top wallets leaving and 3 accumulating, signaling a shift in interest. • SN103 Djinn entered the Risk Radar RED zone with a risk score of 57/100, indicating increased caution. subnetradar.com/alpha/reports #bittensor $TAO #dTAO