August 10, 2026 · Published daily
Subnet News — August 10, 2026
Hippius launched Drive, Score patched a latency-loop exploit, OpenRoboto completed a benchmark upgrade, Beam fixed suppressed task allocation, Verathos shipped v0.1.32, and Poker44 opened miner training tournaments. Developing stories include RedTeam enterprise-scale claims amid scoring failures, a BASE harness exploit awaiting causality checks, Quasar ownership-transition signals, and early Cascade, InfiniteQuant, Targon, and Conjectures milestones. Sixteen subnets produced material reportable signals; the other 112 had no public update strong enough for inclusion.
- Coverage window
- Aug 9, 2026, 10:00 AM – 11:20 AM ET
- Subnets checked
- 128
- Subnets with updates
- 16
What matters first
Top signals
SN75 Hippius Drive goes live
Hippius launched client-side encrypted distributed file storage with user-held recovery credentials and no retrieval egress fees; broad hyperscaler comparisons remain workload-dependent.
SN44 Score patches a latency-loop exploit
Score deployed a tiebreak hotfix and tightened private-track rules while separate model-duplication and detection disputes remain unresolved.
SN80 OpenRoboto completes its benchmark pipeline upgrade
OpenRoboto added granular submission states, live evaluation progress, collision-resistant task IDs, revised ranking rules, and self-service rejection diagnostics.
SN105 Beam fixes suppressed orchestrator allocations
Beam corrected an ownership-grouping bug that deprived some high-performing UIDs of tasks; planned tier and transfer-volume changes are still pending.
SN96 Verathos ships v0.1.32
The release extends proof transport budgets, while frequent probation events and misclassified validator or network failures remain an acknowledged reliability problem.
SN61 RedTeam pairs enterprise claims with active scoring failures
Innerworks reported five enterprise pilots spanning 1B weekly transactions, but miners simultaneously reported dashboard penalties and a broken bot_virus_v1 challenge. The scale and the remediation remain developing.
SN100 BASE fixes WTA weights and finds a harness exploit
Corrected weights deployed on-chain, but a future-token leak that produced an impossible score still awaits the promised causality checks.
SN126 Poker44 opens labeled-data tournaments
Three daily human-versus-AI tournaments now generate immutable micro-sessions for bot-detection training, separated from live evaluation data.
SN24 Quasar ownership transition moves closer
Discussion and conviction locks point to a new control group preparing a handover, but the ownership transfer and timing were not independently confirmed in this pass.
Organized by SN#
Subnet desk
Targon
4090 and 5090 support remains under security review
Targon said RTX 4090/5090 support requires management VMs on Intel TDX- or AMD SEV-capable CPUs while it studies VRAM DMA implications. The team also targeted month-end for 5090-equipped towers; compatibility and timing remain developing until shipped.
Quasar
New Quasar ownership group signals a near-term transition
Team discussion points to a new group preparing to take control through conviction transfers, including a cited wallet with roughly 400K locked alpha. The handover and timeline were not yet independently confirmed on-chain in this reporting pass.
Leoma
Leoma records the largest 24-hour alpha decline
SubnetRadar's August 10 tape showed Leoma down approximately 14.1% in alpha terms, the largest decline on that issue. This is a market observation, not a recommendation or forecast.
Score
Latency-loop hotfix and stricter private-track rules deploy
Score deployed a fix preventing low-score hotkeys from exploiting its latency-loop tiebreak and set a Tuesday deadline for private-track miners to remove external data and OCR-derived speed features.
Duplicate-model and detection disputes remain unresolved
The team flagged miners for duplicate detections and possible model-hash manipulation while a car-wash detection element remained unavailable for more than six hours. The accusations and their effect on final scoring remain developing.
engy
Seven tracked wallets exit about 535 TAO
SubnetRadar's automated chain monitor recorded seven top wallets leaving engy over roughly four hours, led by an approximately 172-TAO sale. The flow is factual market data and does not establish motive.
RedTeam
Dashboard and challenge-scoring failures draw miner reports
Miners reported similarity penalties after formatting fixes and a bot_virus_v1 failure tied to a missing CHALLENGE_WEB_URL variable. The team acknowledged validation changes and required affected solutions to be resubmitted, but did not confirm a replacement-dashboard timeline.
Enterprise pilot reach is reported above one billion weekly transactions
Innerworks and an OpenTensor post said five RedTeam pilots span more than 900M accounts and 1B weekly transactions. The pilots and claimed scale are team-reported and were not independently audited in this window.
RedTeam leads the daily alpha tape
RedTeam was the largest 24-hour gainer on SubnetRadar's August 10 issue at approximately +29.0% in alpha terms. This is market data, not a prediction or recommendation.
conjectures
Conjectures goes live around formal-math competitions
The relaunched subnet said miners are working on two longstanding mathematical problems and published a roadmap for compiler and software-optimization competitions. Launch activity is visible; problem-solve and partnership claims remain self-reported.
Hippius
Hippius Drive launches encrypted distributed file storage
Hippius announced Drive as live, with client-side encryption, distributed storage, user-held recovery credentials, and no egress fees for file retrieval. The launch is confirmed; comparative cost claims depend on workload assumptions.
OpenRoboto
Benchmark pipeline upgrade adds transparent evaluation stages
OpenRoboto completed its backend and database upgrade with granular submission states, live worker progress, versioned task IDs, revised king-of-the-hill rules, and a self-service rejection query.
InfiniteQuant
Swarm trading system enters nine-book live testing
InfiniteQuant reported nine parallel test books across three asset classes and an early 10-win, 3-loss result, alongside a 30-minute cooldown and ratcheting stop-loss. The trial is live, but performance is preliminary and team-reported.
cascade
First pretrained model reaches a reported one-third checkpoint
Cascade said its first model is roughly one-third through pretraining and estimated another three to four weeks at the current miner trajectory. No public progress dashboard was identified, so the milestone remains developing.
Bitcast
Emission rises to approximately 1.05%
SubnetRadar's development monitor flagged Bitcast's emission at approximately 1.05% in the August 10 issue. The figure is a point-in-time network observation and may change block to block.
Verathos
Verathos v0.1.32 extends hard-proof transport budgets
Verathos shipped v0.1.32, aligning proof responses to a 300-second budget and automatically migrating nginx to a 360-second upstream timeout for longer proof-of-work deadlines.
Probation false positives and recovery economics remain open
The team acknowledged validator and transport failures were being misclassified as miner faults, causing frequent probation and near-zero recovery emissions. Reliability changes are promised but not yet demonstrated.
BASE
Corrected winner-takes-all weights deploy on Prism
BASE deployed its corrected WTA mechanism and updated weights on-chain, while adding permanent elimination for designs that go five epochs without a score.
Future-token harness exploit awaits causality checks
A submission achieved an implausible 0.23 bits-per-byte score by reading future tokens. The team identified the exploit and committed to causality validation; the remedy had not yet shipped in the reporting window.
Beam
Orchestrator grouping bug fixed after suppressing task allocation
Beam corrected ownership-group assignments that caused some strong-performing UIDs to receive too few tasks. The team said tier changes and higher transfer volume are next; those follow-on changes remain pending.
Poker44
Poker44 activates miner training tournaments
Poker44 launched three daily human-versus-AI tournaments to create labeled schema-v4.1 micro-sessions for bot-detection training, with hashed public benchmark data kept separate from live evaluation inputs.
No material update found: SN1, SN2, SN3, SN5, SN6, SN7, SN8, SN9, SN10, SN11, SN12, SN13, SN14, SN15, SN16, SN17, SN18, SN19, SN20, SN21, SN22, SN23, SN25, SN26, SN27, SN28, SN29, SN30, SN31, SN32, SN33, SN34, SN35, SN37, SN38, SN39, SN40, SN41, SN42, SN43, SN45, SN46, SN47, SN48, SN49, SN50, SN51, SN52, SN54, SN55, SN56, SN57, SN58, SN59, SN60, SN62, SN63, SN64, SN65, SN67, SN68, SN69, SN70, SN71, SN72, SN73, SN74, SN76, SN77, SN78, SN79, SN81, SN82, SN83, SN84, SN85, SN86, SN87, SN88, SN90, SN92, SN94, SN95, SN97, SN98, SN99, SN101, SN102, SN103, SN104, SN106, SN107, SN108, SN109, SN110, SN111, SN112, SN113, SN114, SN115, SN116, SN117, SN118, SN119, SN120, SN121, SN122, SN123, SN124, SN125, SN127, SN128.
Beyond one subnet
Ecosystem desk
Only 31% of subnet markets rise as 45 subnets remain high-burn
SubnetRadar's August 10 issue recorded 31% of subnet markets rising, 56 new monitored moments, and 45 of 128 reporting subnets at or above 90% miner burn. These are point-in-time network statistics, not financial advice.
The post-v440 field remains in a verification phase
TAO Protocol's weekly pull described incentive retuning, slot renames and ownership changes after the emission gate, while explicitly noting that team figures were not independently audited. The pattern is verified; conclusions about durable demand remain developing.
Audit trail
Coverage notes
Searched
- Public web and X indexing for active SN1 through SN128 by subnet number and registry name during the August 9 10:00 AM to August 10 11:20 AM ET window
- Official team and OpenTensor posts, source-linked Bittensor Discord messages, and public subnet announcements
- SubnetRadar daily issue, per-subnet briefs and chain moments; TAO Daily; TAO Protocol; TaoStats; and official Bittensor surfaces
- Public GitHub repository, release and commit indexing for release, scoring, validator, security, ownership, emission and product-launch claims
- Market-flow, conviction, burn, rising-versus-falling market counts and emission signals, treated as descriptive public data rather than investment recommendations
Public-data gaps
- X indexing was incomplete and some official posts exposed no readable body without login; directly resolved post URLs were retained where corroborating coverage captured the announcement.
- Many primary Discord messages require Bittensor server membership or login; SubnetRadar briefs that link the exact original message IDs were used when direct reading was unavailable.
- Public GitHub search did not reliably surface every subnet repository or release tag, so Discord-announced releases without an attributable public tag were verified only as team deployments.
- Quasar ownership, InfiniteQuant performance, Cascade progress, RedTeam enterprise scale, Conjectures problem solves and future Targon hardware dates remain team-reported and are labeled developing.
- Registry names are moving quickly after slot handovers and rebrands; SN39 was updated to Cathedral from the prior report, while unconfirmed or conflicting slot-name changes were not silently rewritten.
- No item in this report is financial advice. Price, flow, emission, burn and conviction observations describe public data and do not predict future performance.