Bittensor Weekly Update: TAO Price, News & Subnets | 23-02-2026

Big Events in the Bittensor Ecosystem

Upbit Lists TAO With KRW, BTC, and USDT Trading Pairs

South Korea’s largest cryptocurrency exchange has officially added Bittensor to its platform. On February 16, 2026, Upbit announced that TAO trading went live at 4:00 PM KST across three markets: KRW, BTC, and USDT. The announcement also tagged the Opentensor Foundation, signaling a direct relationship between the exchange and the team behind Bittensor’s development.

This is a significant milestone for Bittensor and its community. Upbit consistently ranks among the top global exchanges by daily trading volume, and South Korea is one of the most active retail crypto markets in the world. As a result, the listing gives TAO access to a massive new pool of traders and investors. Historically, tokens listed on Upbit have experienced sharp increases in volume shortly after launch, so the impact on Bittensor’s liquidity could be substantial.

The multi-pair support is also worth noting. By offering KRW, BTC, and USDT markets simultaneously, Upbit makes it easy for both fiat and crypto-native traders to gain exposure to Bittensor. This kind of broad trading access is essential for growing TAO’s presence in Asia and improving overall price discovery across the market.

Subnet Updates

Synth (SN50) Launches $20,000 Hackathon for Bittensor Developers

Synth has announced a hackathon with $20,000 in total prizes. The event challenges developers to build tools that tap into Synth’s probabilistic price data, and it spans three distinct categories.

The first track focuses on options trading tools. Participants are encouraged to create applications that help traders leverage Synth’s predicted price distributions for tasks like strike selection, implied volatility comparison, position sizing, and risk visualization. The second track targets prediction markets, calling for tools that connect Synth’s forecasts to platforms like Polymarket. Example projects include arbitrage scanners, edge finders, and alert systems that flag divergences between market prices and Synth’s probability outputs. The third track is dedicated to equities applications, where builders can use Synth’s volatility forecasts for stocks and indices such as NVDA, TSLA, AAPL, and GOOGL.

Each category offers $5,000 in prizes, split between first ($3,000), second ($1,500), and third place ($500). An additional $5,000 goes to the best overall winner. The hackathon represents a growing trend within the Bittensor ecosystem, where subnets are actively investing in developer engagement to expand their real-world utility. For anyone interested in decentralized AI-powered finance, this is a chance to build on one of the most talked-about data layers in the Bittensor network.

Templar (SN3) Shares TGIF 26 Recap

Templar has published its TGIF 26 community recap. Host DistStateAndMe walked through several major developments, starting with the news that Covenant 72B has officially entered its post-training phase. A full report and the model itself are expected to drop soon.

The Crusades MFU competition also delivered an impressive milestone. Model FLOPs utilization on a single A100 GPU has now reached 66%, more than doubling the 30% achieved during the original training run. This kind of efficiency gain is significant for decentralized training, as it means miners can get more useful compute out of the same hardware.

Perhaps the most forward-looking topic was Heterogeneous SparseLoCo, a training approach that would allow data center GPUs and consumer-grade hardware to work together in the same pipeline. The idea is to turn the entire internet into one distributed data center, which could dramatically lower the barrier to entry for Bittensor miners who don’t have access to enterprise-level hardware.

Beyond Templar’s own progress, the recap highlighted updates from across the Bittensor ecosystem. Grail AI was featured following Erfan’s PULSE presentation at Yannic Kilcher’s event. Basilica (SN39) announced one-click OpenClaw deployments that have already attracted over 100 paying users. DistStateAndMe also shared his plans to build a universal MCP server for Bittensor, aiming to provide a single unified interface for interacting with the network’s expanding toolkit.

Chutes (SN64) Launches ParaClaw

Chutes has released ParaClaw – a tool that lets users deploy a fully functional OpenClaw AI agent in under 60 seconds with a single command. The setup comes with access to over 60 open-source models out of the box, including Kimi K2.5, a 1-trillion-parameter multimodal model that is reportedly outperforming Claude Opus on agentic benchmarks like BrowseComp and LiveCodeBench.

What makes this particularly interesting is the privacy layer. Chutes runs Trusted Execution Environments (TEE), which means GPU operators physically cannot access user data during inference. For an AI agent handling sensitive tasks like emails or file management, that is a meaningful security guarantee baked directly into the hardware.

Click here to learn more.

Bitcast (SN93) Partners With Desearch (SN22)

Bitcast and Desearch have announced an integration that brings structured, real-time X (Twitter) data to Bitcast’s creator marketing platform. The partnership replaces traditional web scrapers with Desearch’s decentralized search infrastructure, giving Bitcast access to validator-level signals for replies, posts, retweets, and campaign engagement.

This matters because Bitcast X, the subnet’s marketing product, is preparing to scale significantly. Reliable real-time data is the bottleneck for any platform that needs to verify creator engagement programmatically, and fragile scrapers have historically been a weak link. By plugging into Desearch’s API, Bitcast gets a more robust and decentralized data pipeline that can handle growth without breaking.

The collaboration is also a good example of how Bittensor subnets are beginning to build on top of each other. One subnet provides decentralized search infrastructure, another uses it to power a decentralized creator economy. That kind of composability between subnets is exactly what the network needs to move beyond isolated projects and toward a more interconnected ecosystem.

Ridges AI (SN62) Delays Product Launch to March 5

Ridges AI has pushed back its product launch by two weeks, with the new target set for March 5. The team says the delay is intentional – they want to get the long-term vision right rather than rush a release that doesn’t reflect the work happening behind the scenes.

When it does launch, users will get access to what Ridges describes as its most autonomous coding agent to date. The goal is a near one-click developer experience, built specifically for decentralized product services at scale. The initial release will still be in beta, but the team expects fast iteration and plans to follow up with a full roadmap and a feature-complete v1.0 shortly after.

SN62 has been widely regarded as one of Bittensor’s strongest contenders for mainstream adoption, and the team led by hobbleabbas appears to be betting that a polished launch will matter more than a fast one. A full roadmap is expected alongside the March 5 release.

ReadyAI (SN33) Ships Website-Level Enrichment

ReadyAI just took a major step toward making the entire web readable for AI agents. The team shipped Webpage Metadata v2 alongside a new high-volume API that enriches full websites rather than individual pages. The update goes mainnet on February 23.

The problem SN33 is solving is straightforward. Search engines break the web into individual pages, which works fine for finding isolated facts but fails when an AI agent needs holistic understanding of a topic. A search for “best skis” might return price rankings from a single page, but it completely misses how factors like waist width affect performance in different snow conditions. That information exists across an entire site – it just hasn’t been structured in a way agents can use.

SN33’s new pipeline changes that. It pushes full websites through the subnet and collects enriched data including tags, named entity recognition, similar page groupings, and cross-page summarization. The end goal is to generate llms.txt files at scale.

The llms.txt standard condenses an entire website’s contents into a single text file that agents and MCP tools can parse without crawling every page individually. Adoption of the standard has been slow because nobody has been producing these files for the broader web. SN33 intends to change that, with the team stating they plan to become the largest llms.txt producer globally.

The subnet already has traction on the data side. Its first open-source dataset, 5000 Podcast Conversations, has surpassed 300,000 downloads on HuggingFace. More open-source releases are on the way as the enrichment pipeline scales up.

Score (SN44) Announces Manako

Score has unveiled Manako, a platform that lets users build production-grade computer vision solutions through natural conversation instead of machine learning expertise. The product is set to launch in Q1 2026 with limited access, and a waitlist is live at manako.ai.

The concept is simple. Users describe what they want to detect or analyze, provide images or video, and Manako handles everything else – from model selection and pipeline optimization to deployment and API delivery. There is no need to design models, configure infrastructure, or understand the underlying architecture. Think of it as an AI coding assistant, but for vision.

Under the hood, Manako runs on three layers: a curated library of vision components that the team continuously optimizes, an AI orchestrator that interprets user intent and assembles the right components into an execution graph, and a clean SDK and API interface that abstracts away all the complexity. The Score subnet feeds directly into this system through ongoing competitions that produce new computer vision models, expanding what Manako can do over time.

Score has spent the past several months building enterprise partnerships across industries including football, cricket, media production, and petrol station monitoring. Those deals validated the technology and proved it could outperform existing market solutions at lower cost. Manako takes that same capability and opens it up to a much broader audience – startups, solo developers, and smaller companies that previously had no way to access advanced computer vision without dedicated ML teams and significant budgets.

New Subnet Launches

Minos (SN107)

Minos is a decentralized genomic variant calling subnet. It turns DNA analysis into an open competition on Bittensor, where miners compete to detect mutations with clinical-grade precision and accuracy is verified by validators rather than taken on trust. The project is still in early stages, with more details expected soon.

Poker44 (SN126)

Poker44 is a decentralized online poker platform that uses Bittensor to power AI-driven bot detection. Game logic runs inside Trusted Execution Environments so private cards and RNG never leave the enclave. Miners on the subnet compete to build the best behavioral detection models, validators score them reproducibly, and the top-performing models are promoted to protect real games in production. Settlement and disputes run fully onchain via EVM smart contracts. The name is a poker inside joke – pocket fours look like pocket aces at a quick glance, a nod to the fact that small details change everything.

TAO Market Update

Price: $171.95-$198.75
Weekly result (7d): -6.50%
Ranking: #46
Market Cap: $1.65B
24h Volume: $76.68M

As of today, TAO trades between $171.95 and $198.75, reflecting a pullback after the sharp rally driven by the Upbit listing earlier in the week. Current market data puts TAO at #46 by market capitalization, with a valuation of $1.65B. TAO recorded $76.68M in 24-hour trading volume, showing continued liquidity and engagement from market participants despite the weekly correction.

Top Gainer Subnet: HashiChain (SN115)

Weekly change: +36.43%

Price: $1.60

Market Cap: $1.91M

Volume (24h): $95.19K

Quick Answers: