Bittensor Toolkit →Find the right tool for your question.

Bittensor Daily News

From the archive
#01 · SN105 · Beamverified
★★★★★

SN105 Beam connects data-transfer receipts to Chainlink CRE

Beam announced its Chainlink Runtime Environment integration on November 13, and Chainlink independently listed Beam among its new integrations that day. The connection gives applications a way to receive verifiable on-chain records of data transfers. Beam says transfer reports are already recorded as Chainlink-attested receipts on Base mainnet, with the first covering 100 GB. Smart contracts initiating the transfers are the next planned stage, not a capability the announcement says is already live. A delivery receipt could help an application decide when to start a training job or advance a document workflow; it does not prove the contents of a dataset are correct. This is dated catch-up coverage of a substantive integration. The relationship is confirmed by both parties. The first-transfer size and Base deployment are attributed to Beam; this review did not independently replay the transfer or audit its receipt contract.

Why 4 stars: Four stars: a counterparty-confirmed integration connects subnet infrastructure to smart-contract workflows. Three understates the new integration scope; five is not established by a first transfer and proposed use cases.

#02 · SN80 · OpenRobotoverified
★★★★★

SN80 OpenRoboto opens an agentic robotics task-building track

OpenRoboto announced its Agentic Robotics Track on November 12. The new task network is intended to test how AI models observe, reason and act in simulated physical environments. The available Track A asks miners to create robot-task packages with a scene, success conditions and a reference demonstration. The team describes checks for validity and reproducibility before reference models are evaluated. Track B, where miners run assigned evaluations and return execution evidence, is explicitly marked coming soon. For builders, this expands the subnet's evaluation infrastructure around tasks that can distinguish model capabilities. It is a new track launch rather than a report of a routine competition winner. The team announcement establishes availability as reported by the operator; we have not run the task pipeline or independently verified model-performance claims.

Why 4 stars: Four stars: a substantive new robotics evaluation track changes what miners can contribute. Three would fit routine results within an existing competition; five is not supported by demonstrated broader adoption.

#03 · SN54 · Yanezverified
★★★★★

SN54 Yanez reports sales growth and estimated ARR above $500,000

In a November 12 Ventura Labs interview, Yanez CEO Jose Caldera said sales had doubled across Q3 and October, bringing estimated annual recurring revenue to slightly above $500,000. He explicitly described the ARR figure as an estimate because some products are new and their recurring revenue is still difficult to measure. It is not a statement that the company collected $500,000 during those four months. Caldera also said Yanez had signed a large social-network customer for compliance and risk technology, but could not identify it under an NDA. The customer identity and contract value are not established by the interview. The distinction between product lines matters: Caldera said many larger customers use Yanez's original financial-crime and regulatory-testing products, while proof of humanity and uniqueness is a newer line launched during the summer. These company-wide figures should not be presented as revenue generated entirely by SN54's newer product or as payments to token holders. This is dated catch-up coverage of an attributed commercial update, not independently audited financial results.

Why 4 stars: Four stars: a concrete company-reported commercial milestone and new customer materially extend the adoption picture. Three understates that commercial development; five is not established by estimated company-wide ARR and an unnamed customer.

#04 · SN4 · Targonverified
★★★★★

SN4 Targon expands SDK access across its compute products

Targon announced an SDK update on November 12 covering virtual machines, dedicated bare-metal servers and sandboxes. Its current SDK v4 repository documents typed interfaces for managing these workloads from code. There is an important language distinction: Go and Python support all three workload services, while TypeScript currently supports Sandboxes only. The repository is more specific than the social announcement's wording about language coverage. It also says the previous Rental and Volume APIs are unavailable in v4 with no current replacement. For developers, the update brings more of Targon's compute management into a common toolkit, including sandbox files, execution, forks and templates. Existing integrations should check the documented migration path rather than assume every older API remains compatible. This review checked the published interface documentation, not a paid compute deployment.

Why 3 stars: Three stars: a meaningful SDK expansion and migration change for existing compute users. Four would require a more consequential new service or integration beyond developer access to existing products.

#05 · SN121 · sundae_barverified
★★★★★

SN121 Sundae highlights Scout’s context-aware agent handoff

Sundae described Scout on November 15 as a way to turn a user's description of work into an agent handoff. The team says Scout remembers the user's role and preferences, then launches an agent suited to the job; its follow-up invites users to try it with a free account. The update focuses on choosing and briefing an agent rather than requiring the user to select one manually. It is a team-described product capability, not evidence of task success rates or paying adoption. No account, credential vault or automated task was used in this review.

Why 3 stars: Three stars: a useful agent-selection and context feature with bounded demonstrated scope. Four is not established by the short product announcement alone.

#06 · SN116 · Carbonverified
★★★★★

SN116 Carbon reports its first miner payment after launch

Carbon said on November 15 that it had a new competition leader and had paid its first miner. The announcement follows its November 14 mainnet launch and first motor-design challenge, already covered in the previous edition. This is an operational follow-up to that launch, not another launch announcement. The post does not provide a payment amount or transaction reference, so the payment is reported as the team's statement rather than an independently reconciled chain transfer.

Why 3 stars: Three stars: the first reported miner payment is a meaningful operational milestone, while a new competition leader remains within the routine-competition ceiling.

#07 · SN127 · Zilsverified
★★★★★

SN127 Zils gives miners clearer feedback on rejected submissions

Zils published two optional releases on November 11 that help soup miners understand why their submissions did not win a seat. The private round report shows the stage reached, the reason a submission stopped, and its scores against the required thresholds. A miner signs the request with their hotkey and receives information about their own submissions only. Reports apply to rounds graded on the new release or later. The follow-up release routes the command through the hosted API by default, so miners do not need to supply a validator address. The command verifies the validator's signed response. The relay can read that response, so private here does not mean hidden from the relay; miners can still request it directly from a validator. The earlier release also retries unanswered queue lease renewals to reduce dropped jobs. The release notes say scoring, seats and the public feed are unchanged. This is a dated operational update, not a new reward formula.

Why 3 stars: Three stars: useful diagnostics and job reliability changes affect participation. Two understates the new feedback capability; four would overstate a release that leaves scoring and rewards unchanged.

#08 · SN64 · Chutesverified
★★★★★

SN64 Chutes lists Tencent Hy4 with separate cached-input pricing

Chutes' public model catalog checked on November 15 lists Tencent Hy4 with a 1,048,576-token context window. Its quoted USD rates per million tokens are $0.525 for fresh input, $0.025 for cached input and $1.5625 for output. For an application repeatedly sending the same instructions or reference material, the distinction is practical: input billed as a cache read costs less, while new input and generated answers have separate rates. The listing does not establish how often a particular workload will qualify for the cache rate. This is verified catalog availability, not evidence of a Tencent partnership or a measured performance advantage. The model was already discussed publicly before today's report; it is included as a missed product update, not a November 15 launch. No paid inference request was submitted in this review.

Why 3 stars: Three stars: a meaningful catalog addition with useful pricing details. Two understates the available model choice; four is not warranted by hosting alone.

#09 · Ecosystemverified
★★★★★

Boost Run filing separates GPU rental growth from network rewards

Boost Run's quarterly filing reports $39.87 million in revenue for the three months ended September 30, 2026. Compared with the same quarter a year earlier, GPU leasing revenue increased by $35.59 million, partly offset by a $2.46 million decline in blockchain-reward revenue. Those are changes in revenue, not the absolute revenue of each segment. The company describes providing compute to Bittensor and Aethir in exchange for TAO and ATH, recording those rewards at fair value and converting them to cash or using them for expenses shortly afterward. The combined reward decline is not broken out by network in the cited discussion, so it cannot be described as a $2.46 million decline in Bittensor alone. For readers tracking commercial demand, rental contracts and protocol rewards are different payment flows. This filing does not identify the Bittensor subnets supplied or establish how much end-user demand those rewards represent.

Why 3 stars: Three stars: original issuer accounting gives useful context for compute economics. Two understates the quantified disclosure; four is not established without a more direct subnet-specific development.

#10 · SN94 · Cathedralverified
★★★★★

SN94 Cathedral separates agent failures from infrastructure errors in its test report

Cathedral's November 14 technical article revisits a November 6 test of 1,000 AI-agent attempts in isolated sandboxes. The public run card reports 424 solved and sealed, one solved with its seal unproven, 557 failed and 18 infrastructure errors. The distinction matters for evaluation: a machine failing before grading should not be counted as the model giving a wrong answer. The card lists the model, execution setup and result categories. It also exposes important limits that the short social post does not convey: the summary uses a throwaway prototype signing key, the advertised production verification command does not exist yet, and the complete run folder needed by the prototype verifier is not yet published. The page says receipts establish execution and output, not that the grader itself is correct. This review read the team's article and run card; it did not independently verify the receipt chains. The result is a bounded demonstration of execution reporting, not proof of general model performance or completed adoption by other subnets.

Why 3 stars: Three stars: a substantive test report makes execution evidence and its limits inspectable. Two understates the technical detail; four is not established by a bounded test with prototype verification artifacts.

#11 · SN70 · Fugalverified
★★★★★

SN70 Fugal fixes proof-collection delays that could penalize responsive miners

Fugal's November 12 v0.3.2 changelog describes a validator fix for proof requests getting stuck behind unresponsive miners. The previous connection pool could be occupied by more than 100 endpoints that did not answer, delaying requests to miners whose proofs were ready. The team reports this happening on Finney and says a missed proof could reset a head's burn-in before it completed. The update changes proof collection so those unanswered connections no longer block the rest of the field. The documented timeout, queried endpoints, scoring and consensus digest remain unchanged. This is a released validator fix, not another announcement of the v2 preparation covered on November 10. The change is relevant operational context, but it does not establish why Fugal's token fell. The November 15 market snapshot showed a 12.63% 24-hour decline against TAO at 17:26 UTC; the original team Discord discussion remains inaccessible, and no causal explanation has been verified.

Why 3 stars: Three stars: the fix addresses an operational defect affecting miner participation and burn-in. Two understates that effect; four would overstate a validator-only fix with scoring and consensus unchanged.

#12 · SN87 · Provenonceverified
★★★★★

SN87 Provenonce shifts its validator and miner emission split

A finalized November 14 transaction changed Provenonce's five-slot emission allocation at block 9505951. The validator share rose from approximately 41% to 49%, while the miner share fell from approximately 41% to 33%. The owner share remains approximately 18%, and the burn and recycle slots remain zero. The exact integer vectors are retained in the linked evidence. This changes how the subnet's allocated emissions are divided; it does not mean the subnet receives more network emissions. Validator distributions still interact with the protocol's Root allocation rules, and displayed staking yield can lag as epochs settle. The review verified successful execution and the before-and-after state, rather than inferring a change from an APY chart.

Why 3 stars: Three stars: an executed eight-percentage-point redistribution matters to miners and validators. Two understates the economic effect; four is not established by this bounded allocation adjustment.

#13 · SN8 · Vantaverified
★★★★★

SN8 Vanta records ten qualifying alpha-burn transactions

The finalized review found ten successful SN8 alpha-burn events between November 14 at 17:55 UTC and November 15 at 17:24:36 UTC. Each individually exceeded the public archive's 100-alpha threshold. Together they removed 1,967.787795218 alpha. These are alpha units, not TAO or dollars. The events are recorded as direct alpha burns, not buy-and-burn transactions: the reviewed events do not establish a matching TAO purchase. The total excludes smaller events, automatic emission withholding, fees and recycling. Each qualifying transaction is retained separately in the Burns archive with its block, event index and amount. The chain establishes the burn execution, not who beneficially owns the originating wallets or why the tokens were burned.

Why 3 stars: Three stars: verified qualifying token destruction is a meaningful supply update. Two understates the confirmed transaction series; four would require greater demonstrated consequence than the amounts alone establish.

#14 · SN76 · Ormasverified
★★★★★

SN76 Ormas enables Gamma, with no service registered at the cutoff

Ormas enabled Gamma in a successful transaction at finalized block 9507253 on November 14. The November 15 snapshot at block 9513318 shows nine Gamma-enabled subnets, including Ormas. Enabling the feature and offering a usable service are separate steps. Ormas had no service entry in the on-chain Services registry at that cutoff. This should not be presented as a customer launch, paid adoption or proof that someone bought and consumed a service. The linked snapshot preserves the enabled flag and empty service list.

Why 3 stars: Three stars: a verified new Gamma participant changes the adoption roster. Two understates the concrete enablement; four is not established without a registered service or substantive launch.

Stars indicate editorial priority, not investment quality. Routine competition updates receive at most three stars. Shizzy videos receive three; substantive non-video Shizzy coverage receives at least four. Scheduled events remain plans.

Sources & coverage notes
04

Audit trail

Coverage notes

Beam connects data-transfer receipts to Chainlink CRE. OpenRoboto opens a robotics task track, Yanez reports commercial growth, and follow-up coverage adds developer tools, compute disclosures, Gamma enablement and verified emission and burn events.

Coverage: Nov 14, 2026, 10:58 AM – Nov 15, 2026, 12:39 PM ET

Searched

  • Direct original announcements from Beam, Chainlink, OpenRoboto, Targon, Sundae and Carbon; official Targon SDK documentation.
  • Direct bounded reviews of the required ecosystem X profiles, active partner Injective, fund principals and discovery outlets; actual signed-in Grok discovery, attention and omission queries.
  • Follow-up: 128 configured website rows attempted (106 HTTP 200, 19 without a configured website, three access errors); TAO Daily, SubnetRadar, Tao Outsider, AlphaGap, Taosis and 1t1b discovery pages; original Yanez interview segment, Zils releases, Chutes catalog and Boost Run filing.
  • Full finalized event interval 9505682-9513318, execution checks for qualifying burns and emission changes, lifecycle events and Gamma enablement/service snapshot.

Public-data gaps

  • Full seven-day social, media, fund-principal and team Discord coverage remains incomplete. Roster inclusion is not a claim that every subnet was fully reviewed.
  • Discord originals, complete seven-day team/fund/media review and Gamma provider verification remain incomplete. The chain interval is reviewed through finalized block 9513318; later events are outside that cutoff.
Browse all 100 editions

Conference news · September 30, 2026

Exploit Conference Catch Up · Day 1

The teams, the ideas and the practical takeaways from Exploit Summit’s first day in Montreal. From Gamma and phone-based AI to robotics, science and the business of running a subnet.

Catch up on Day 1

Conference news · October 1, 2026

Exploit Conference Catch Up · Day 2

Catch up on Exploit Summit’s second day: Carbon’s Pitchtensor win, SN107 Minos’ research app, agent controls, robotics data, inference partnerships and the decisions behind building a subnet business.

Catch up on Day 2