Subnet focused articles visible today.
Subnet Research
Every subnet is a market, a product and a bet on useful work.
Tao Outsider researches Bittensor subnets through proof, not personality. Product, incentives, code, liquidity, emissions, flows and failure modes all matter.
Evidence fields required before a thesis feels serious.
Posts, GitHub, product pages, dashboards and market structure.
The method
The field lens
Product surface
What can a real user, customer, miner, developer or institution actually touch today?
Incentive design
What behavior is rewarded, how is it measured and where could miners optimize the wrong thing?
Market structure
How deep is the pool, how painful is slippage and what does price say after flows are separated from noise?
Validator quality
Who is judging the work, how stable are weights and what evidence exists beyond brand names?
Code and shipping
What changed in GitHub, docs, releases and public infrastructure since the last story was told?
Narrative risk
Does the subnet have proof, or only a beautiful explanation that investors want to believe?
Research queue
What should be watched next
Revenue, inference demand and product traction.
Long context training and competitive model improvement.
Hardware, confidential compute and real world deployment.
Risk intelligence and the first steps of native finance.
Validator allocation, reinvestment and capital circulation.
Tracked files
Subnets already on the desk
SN64 Published analysis
Chutes
Compute and inference
Chutes is one of the clearest Bittensor stories when product distribution, revenue claims and infrastructure costs are read together.
Open research SN23 Published news report
Trishool
AI safety
HaloGuard matters because it turns miner pressure into a benchmarked AI safety artifact that can be checked outside subnet culture.
Open research SN81 Published news report
Reliquary
Training market
Reliquary is worth watching because miners search for prompts near a model learning frontier instead of only producing generic activity.
Open research SN91 Published launch context
Cascade
Time series models
Cascade gives Bittensor a time series model angle, but the next proof is public code, miner onboarding and measurable validation.
Open researchResearch artifact
What a subnet note must contain
A subnet note is not a vibe check. It needs a claim, evidence, numbers and a clear invalidation path.
- One paragraph thesis
- Evidence table
- Main product proof
- Incentive failure mode
- Liquidity and slippage read
- What would change our mind
Published research
Latest subnet intelligence
Analysis
I expect TAO to revisit its all-time high
My Bittensor thesis rests on broader investor access, paying subnet customers and a dependable protocol. Here is what could support a TAO return to its high.
Analysis
IOTA wants Bittensor builders to bring the workload, not manage the machines
Macrocosmos outlines an IOTA SDK for distributed training on Bittensor SN9. What developers would control, and what remains a preview.
Analysis
Bittensor's Cascade puts forecasting AI within reach with Ephemeris
Cascade SN91 brings six forecasting models into one API with Ephemeris, opening a practical route into time series AI for developers and agents.
Analysis
Bittensor SN46 dispute: Dan says Instant is leaving, Seby pushes ahead
Two Instant accounts and conflicting plans for Bittensor SN46. Lily Venice follows the dispute, the real estate pivot and what it means for the TAO ecosystem.
Analysis
KubeTEE adds image and video generation to its Bittensor gateway
KubeTEE SN90 reports adding image generation and video with audio to its public gateway, expanding the workloads served by its miner clusters.
Analysis
ORO changes the test for Bittensor shopping agents
ORO's September 17 update replaces ShoppingBench with ORO Bench, giving SN15 agents shared qualifying tasks and separate checks for seven task families.
Analysis
Subnet Metrics separates Bittensor emission settings from observed TAO flows
Subnet Metrics puts Bittensor's TAO emission switches beside observed network flows, with separate treatment of alpha rewards and historical coverage.
Analysis
TaoBubbles gives Bittensor a market map worth exploring
A hands-on look at TaoBubbles, its Bittensor subnet map, filters and market tables, plus six features we would like to see for deeper research.
Subnet Deep Dive
Conjectures SN66 publishes an Erdos 96 proof package
Bittensor's Conjectures SN66 publishes a paper, Lean source and review record for a claimed negative answer to Erdos problem 96.
Subnet Deep Dive
Reliquary SN81 hardens proof throughput and restart recovery
Seven merged Reliquary changes target proof batching, checkpoint races, restart-safe weights and stalled duplicate prompts without proving a public V1 launch.
Subnet Deep Dive
Blockmachine SN19 v0.2.13 stops missing data from becoming a burn vote
Blockmachine's validator preserves the last nonzero on-chain weight vector when startup failures hide traffic data, instead of treating blindness as a 100% burn signal.
Subnet Deep Dive
Lium SN51 ships a provider control plane across three releases
Lium v0.0.43 through v0.0.45 add workspace isolation, node-status tooling, account audit logs and a one-command provider registration path.
Subnet Deep Dive
SparkInfer adds LM Studio and Ollama APIs around one inference core
SparkInfer has merged LM Studio and Ollama-compatible REST surfaces that translate into its existing generation handlers without creating a second engine.
Subnet Deep Dive
Pareton SN10 adds an SGLang path for Qwen3.8 FP8 campaigns
Pareton SN10 has merged SGLang-native build, serving and scoring support with a pinned Qwen3.8-27B FP8 campaign path, but launch remains an operator step.
Subnet Deep Dive
Inference Labs SN2 adds a guard against owner-only weight commits
Inference Labs release 14.14.3 holds prior Bittensor weights when every non-owner miner receives zero, closing an owner-only normalization path.
Subnet Deep Dive
Lium SN51 v0.0.39 gives GPU renters a machine-readable path
Lium v0.0.39 adds JSON output, documented error handling and a GPU-spec rental flow that selects the cheapest eligible listed node.
Subnet Deep Dive
Verathos SN96 v0.2.1 packages cross-machine Sleipnir serving
Verathos SN96 v0.2.1 adds an optional cross-machine serving package with worker joining, authenticated weight retrieval and startup readiness checks.
Subnet Deep Dive
OpenRoboto SN80 says 85% of track emissions now target real robots
OpenRoboto now documents a 15% simulation, 42.5% pi0.5 and 42.5% LingBot-VLA 2.0 allocation across three separate robotics tracks.
Subnet Deep Dive
Lium SN51 patches a pod-rental retry that could duplicate a GPU order
Lium's open SDK now checks whether a pod was created before retrying an ambiguous rent request and reuses one idempotency key for the bounded retry.
Subnet Deep Dive
Zeus SN18 cuts challenge weights after eight consecutive penalties
Zeus v2.1.4 excludes established miners from an affected challenge's weight calculation after eight consecutive penalties, with a four-week grace period.
Analysis
Fugal v0.2.0 repairs its TEE verifier before a confirmed Finney launch
Fugal reports closing five TEE verifier exploit classes and three weight-blocking defects in a pre-Finney Bittensor LLM router release.
Subnet Deep Dive
SparkInfer v0.5.5 adds video input and repairs concurrent requests
SparkInfer v0.5.5 brings bounded video input, reusable sequential prefixes and a concurrency reliability fix to Gittensor's inference runtime.
Subnet Deep Dive
Leadpoet SN71 merges miner-built sales agents after an isolated testnet run
Leadpoet SN71 has merged a Python-agent submission workflow after an isolated testnet exercise, while production intake, rewards and chain writes remain disabled.
Subnet Deep Dive
Ridges SN62 v0.3.0 makes coding-agent competitions first-class
Ridges SN62 release v0.3.0 makes uploads, scoring policy, scheduling, leaderboards and emission allocation aware of multiple coding-agent competitions.
Subnet Deep Dive
Score Studio previews SAM3 tennis tracking before its September launch
Score has previewed quantized SAM3 tracking inside Score Studio, an upcoming computer-vision product connected to Bittensor subnet 44.
Subnet Deep Dive
TPN SN65 opens TPN-003 to fit a 12B model below 8 GiB
True Performance Network has opened TPN-003, a Bittensor competition to compress a Gemma4 12B model below an 8 GiB RAM ceiling across five benchmarks.
Subnet Deep Dive
Bitsec SN60 launches Sentios for continuous adversarial code review
Bitsec has launched Sentios, a continuous code-review service built around competing AI security agents, public pricing and repeat scans after meaningful code changes.
Analysis
Chutes SN64 trace maps a year of changing LLM workloads
A Chutes SN64 research trace covers 6.12 billion LLM requests and shows why prefix-aware routing matters as serving workloads change.
Analysis
TaoFlows adds subnet swaps, watchlists, alerts and portfolios
TaoFlows has expanded its Bittensor flow terminal with subnet swaps, watchlists, on-chain alerts and a portfolio dashboard.
Subnet Deep Dive
Bitcast SN93 turns its creator engine into an application API
Bitcast SN93 now documents a versioned application API that lets specialist creator products use its campaign and scoring engine.
Subnet Deep Dive
Synth SN50 adds Ultra scores to a conditional validator reward blend
Synth SN50 code can blend externally supplied Ultra scores into validator rewards as a fourth competition, but the path stays off without a provider.
Subnet Deep Dive
SOMA SN114 opens its context compressor to GitHub Copilot users
SOMA SN114 has opened a context-compression product for GitHub Copilot, with roughly 10% token savings claimed for DeepSeek V4 Pro.
Subnet Deep Dive
Cacheon SN14 gives its newest lineage champion an 80% reward floor
Cacheon SN14 says its new emissions policy settles every pass and gives the newest champion at least 80% of a lineage's existing pooled credit.
Subnet Deep Dive
Gittensor SN74 shifts inference rewards toward verified output tokens
Gittensor SN74 code now measures verified output tokens for serving rewards as its RTX 5090 model reaches 303,714 Hugging Face API-reported downloads.
Subnet Deep Dive
OpenRoboto SN80 moves its robotics tournament to LingBot-VLA 2.0
OpenRoboto SN80 has made LingBot-VLA 2.0 the baseline for its current simulation tournament, with promotion tied to project-defined scoring conditions.
Analysis
Bittensor's Red Queen: The Upside Below the Top 32
The Top 32 emission gate does not end the race. It may create Bittensor's most asymmetric subnet research opportunity below the line.
Subnet Deep Dive
Compelle SN82 code makes its AI debate title fight harder to win
Compelle SN82 repository code scores title fights with complete seat-balanced pairs, a win margin and paired significance while rejecting prompt extraction.
Subnet Deep Dive
AdTAO SN21 makes its allocation curve tail switchable and dated
AdTAO SN21 published code and a preview tool for changing allocation-tail decay on an effective date while recording the curve used in its audit.
Subnet Deep Dive
Harnyx SN67 changes how research agents reach a novel score
Harnyx SN67 code adds fast-mode task routing, a second novelty challenge and clearer failure attribution to its Bittensor scoring path.
Subnet Deep Dive
InfiniteQuant SN89 builds signed-points scoring; the switch stays off
InfiniteQuant SN89 code prepares a 30-day signed-points ledger for wins, losses and washes, while HF_POINTS_FROM remains zero.
Subnet Deep Dive
MVTRX SN79 0.6.0 prepares exchange agents while mainnet stays simulation-only
MVTRX SN79 0.6.0 gives programs a shared simulation and exchange interface; the exchange mechanism remains localnet-only.
Analysis
ORO SN15 code moves reasoning scoring to fail closed
ORO SN15 code makes incomplete reasoning-judge coverage invalidate an evaluation and adds a public weight fallback; validator-wide deployment is unverified.
Subnet Deep Dive
Verathos SN96 v0.1.42 tightens proof paths and account access
Verathos SN96 v0.1.42 packages hard audit reliability, wallet accounts, API keys and payment access while Qwen3.6 35B awaits separate activation.
Subnet Deep Dive
Ditto SN118 v0.133.0 stages node-scoped memory screening capacity
Ditto SN118 v0.133.0 adds per-node screening controls and a staged dedicated-fleet design with GCE overflow, while the prepared host remains unenrolled.
Analysis
How Bittensor Rewired Subnet Emissions, Runtime 421 to 450
A source-bound investigation of Bittensor subnet emissions, the V440 gate, MinerBurned, Root Reborn, Conviction and beta pricing through Runtime 450.
Subnet Deep Dive
Bittensor SN25 draws 2,730 TAO as Keith Singery teases launch
Bittensor Subnet 25 recorded about 2,730 TAO of net inflow over seven days while Keith Singery revived expectations around its next project.
Subnet Deep Dive
Opentensor reports RedTeam SN61 bot detection rose sharply
Opentensor reports that RedTeam's bot-detection rate rose from 74.3% to 99.5% across 11 months of live attacks. The useful question is how the measurement was built.
Subnet Deep Dive
Vidaio SN85 launches Sentinel for Bittensor security video workflows
Vidaio says Sentinel is live as a surveillance-focused version of its Bittensor-native video compression product for security footage and CCTV workflows.
Subnet Deep Dive
Score SN44 puts its 100ms edge-vision test live on Bittensor
Score says SN44 now checks vision miners hourly against random challenges, rewarding models that stay accurate and fast enough for its sub-100ms target.
Subnet Deep Dive
Glyph SN117 puts Bittensor compression into public view
Glyph SN117 says its lossless neural compression ratio reached 9.12%, while public validator logs let Bittensor readers inspect the claim.
Subnet Deep Dive
Yanez SN54 maps revenue into its Bittensor alpha loop
Yanez says Proof of Humanhood revenue will convert into SN54 alpha, connecting its identity product to a new Bittensor incentive loop.
Subnet Deep Dive
404 SN17 brings Bittensor 3D tools into creator apps
404 SN17 is exposing Bittensor 3D generation through web, Discord, Blender and Unity tools, moving the subnet closer to creator workflows.
Analysis
Jason Calacanis says Bittensor cut token costs 95%
Jason Calacanis says a Bittensor inference subnet cut his token costs 95%. Chutes SN64 and Engy SN53 are both in the discussion, but he never named the subnet.
Subnet Deep Dive
NIOME SN55 pushes the Bittensor DeSci case into synthetic genomes
NIOME says Bittensor SN55 can support gene-therapy research with realistic synthetic genomes, a DeSci story that depends on downstream validation.
Subnet Deep Dive
Kraken Listings says Lium SN51 is live for Bittensor market access
Kraken Listings says Lium SN51 is live, giving Bittensor users a clearer market access surface for a GPU rental subnet.
Subnet Deep Dive
ChronoLLM SN38 tests time bounded LLMs on Bittensor
ChronoLLM SN38 is building a Bittensor competition around chronologically consistent LLMs, with recent code work on scoring, anti copy checks and validation.
Subnet Deep Dive
ReadyAI SN33 points revenue toward coding data
ReadyAI says SN33 had 3x revenue in June and is moving toward coding data for AI agents, but the revenue claim remains project reported.
Subnet Deep Dive
Enigma SN63 raises the RSA challenge after one year on Bittensor
Enigma SN63 marked one year on Bittensor with new RSA challenge milestones, project reported submissions and a 500 bit RSA prize target.
Subnet Deep Dive
Chutes SN64 says Parallax trained without blocking GPUs
Chutes says Parallax trained a recurrent model across distributed GPUs with a stated 0.6 percent quality gap versus centralized training.
Subnet Deep Dive
Yanez SN54 previews Proof of Humanhood after launch claim
Yanez says Proof of Humanhood went live on July 1, with client integrations and more details expected on Novelty Search.
Subnet Deep Dive
Forge and Score connect Bittensor lending to SN44
Forge says it partnered with Score as it builds Bittensor native money markets around TAO and subnet collateral.
Analysis
SOMA SN114 turns context cost into a Bittensor target
SOMA says it can cut long context agent cost, while fresh GitHub work shows SN114 moving its screening mechanism forward.
Analysis
TAO Wallet extension brings subnets closer to users
TAO.com says its rebuilt Chrome extension is focused on subnet trading, moving more Bittensor actions into the wallet.
Subnet Deep Dive
Babelbit SN59 moves toward paraphrasing
Babelbit says SN59 is expanding from real time interpretation toward paraphrasing, accent adaptation and same language speech changes.
Analysis
Bittensor DeSci: Claims, NOVA, Minos and NIOME
Claims SN111, NOVA SN68, Minos SN107 and NIOME SN55 show how Bittensor DeSci can turn scientific work into measurable subnet markets for AI and biotech.
Subnet Deep Dive
Chutes SN64 brings dFlash model news to Bittensor
Opentensor says Chutes SN64 has its first in house dFlash model live, with Qwen3 32B reported at about 50 percent higher throughput.
Subnet Deep Dive
Score SN44 prepares a speed test for Bittensor vision miners
Score says a Monday update will require models to be accurate and fast, with top miners already under 60ms before the 100ms bar.
Subnet Deep Dive
Blockmachine SN19 opens free public RPC access for Bittensor
Blockmachine says its free Bittensor RPC endpoint is live with no account or API key required. The story is developer access.
Subnet Deep Dive
AdTAO SN21 puts revenue claims into Bittensor news
AdTAO says its June update added franchise customers, minimum ARR and miner payouts. The useful Bittensor story is proof of value.
Subnet Deep Dive
Bitsec SN60 brings code security agents to Bittensor
Bitsec SN60 turns AI code security agents into a public contest with sandboxed evaluation, cost-bearing miners and downloadable evidence.
Subnet Deep Dive
Zeus SN18 makes Bittensor forecasts verifiable
Zeus SN18 published a public archive of historical weather forecasts with blockchain proof, giving Bittensor a cleaner story for energy market data.
Subnet Deep Dive
Chutes SN64 turns AI power into a Bittensor story
Chutes is using the cost of AI power infrastructure to frame why decentralized training and compute markets matter for Bittensor.
Subnet Deep Dive
Reliquary SN81 opens a Bittensor training market
Reliquary SN81 is live on Bittensor with a mechanism where miners search for prompts at a model's learning frontier and submit verified rollouts for training.
Subnet Deep Dive
Trishool SN23 gives Bittensor an AI safety story
Trishool says HaloGuard 1.0 reached leading prompt safety results among evaluated open weight guard models. The bigger story is the miner loop behind the model.
Subnet Deep Dive
Kraken roadmap puts Bittensor subnet tokens in view
Kraken's public listings roadmap now shows seven Bittensor subnet alpha tokens, giving dTAO a new exchange visibility signal.
Subnet Deep Dive
SN91 Cascade: Bittensor's time series subnet
SN91 Cascade aims to train time series foundation models on Bittensor through synthetic data, miner competition and Bitstarter's first launch slot.
Subnet Deep Dive
Quasar SN24 puts long context on the board
Quasar Preview gives SN24 a public checkpoint, an 18B MoE architecture and a visible path for decentralized long context training.
Analysis
Root Reborn shipped in Bittensor Runtime 441, but the strategy gate remains
Root Reborn code is packaged in Bittensor Runtime 441. Validator-directed subnet baskets remain gated until governance enables root weights.
Subnet Deep Dive
Targon Tower Pro makes compute physical
Targon SN4 is turning Bittensor compute into a hardware story with Tower Pro, confidential GPUs and an earning mode for idle capacity.
Subnet Deep Dive
Chutes makes the revenue story harder to ignore
Chutes SN64 is building a stronger Bittensor case through product distribution and reported revenue, while Parallax adds a new training angle.
Subnet Deep Dive
Minos SN107 puts DeSci on a measurable track
Minos SN107 is turning genomics into a Bittensor workload with synthetic genome generation, variant calling and agent ready subnet context.
Subnet Deep Dive
Chutes SN64 is becoming easier to test
Chutes SN64 is becoming easier to test as TypingMind and VS Code bring its inference path closer to real users.