Bittensor Daily News
Latest editionSubConnect estimates $70M annualized subnet revenue; Barry Silbert highlights the growth
On November 4, SubConnect raised its estimate of annualized revenue across revenue-generating Bittensor subnets to $70M, from $50M a month earlier and $35M two months earlier. That is 40% growth in the latest monthly comparison and a doubling across two months. Its graphic dates the snapshots September 1, October 1 and November 1. At 7:44 AM Eastern, Barry Silbert quoted the report and highlighted the 40% growth. This measures an estimated annualized run rate, not $70M already earned, audited ecosystem accounts or profit. SubConnect explicitly says sustaining growth, margins and reduced emission dependence remain the test. Its $100M-plus year-end figure is an expectation, not a completed milestone.
Around Bittensor
19 more storiesSecurity patch merges into release-487; mainnet activation is separate
On November 3, Subtensor PR3299 merged into release-487. It tightens delayed-proxy authorization and basket-dividend minting against stale permissions and short-lived price manipulation. This is a verified code merge, not proof of a mainnet upgrade: today’s finalized review used runtime 486. No exploitation or user losses were established in this review.
SN4 Targon’s dashboard shows $1.159M in trailing 30-day platform revenue
SN4 Targon’s public revenue dashboard, checked November 4, shows $1,158,560 of combined platform revenue over the trailing 30 days, alongside 15,910 total users and 8,183 new workloads in that period. The chart separates VM, Contracts, BM and Sandboxes. These are operator-published operating metrics, distinct from emissions, token valuation and profit. A secondary report puts October revenue near $1.2M; this review verified the live trailing-window figure and does not substitute it for a reconstructed October calendar total. The dashboard’s network snapshot is dated November 4 at 17:03 UTC; the revenue window has its own 30-day definition.
SN9 IOTA announces open IOTA Decide 4B weights and distributed-training details
Macrocosmos announced SN9 IOTA’s first public model on November 4 at 12:11 PM Eastern. The IOTA Decide 4B v0.4 repository exposes Apache-2.0 weights, a model card and a loader; its file history predates today’s announcement. The text-only model takes a state and a typed question and returns probabilities for the available answers, using one forward pass per question rather than generating a chat response. The team describes fine-tuning Qwen3.5-4B across 16 independently hosted RTX A4000 GPUs using streaming DiLoCo. That training account was not independently reproduced. Builders can inspect and self-host the artifact, but v0.4 requires the bundled letter_decide.py loader: ordinary chat-generation serving changes its behavior. The card reports a 78.6 JevBench composite and says leaderboard listing is pending because the newer question draw has not yet been equated. Accordingly, this report does not endorse the announcement’s top-ten or Pareto-frontier ranking as an established like-for-like result.
Talisman announces detection and withdrawal of TAO at mapped EVM addresses
On November 4, Talisman explained the distinction between ordinary EVM-key accounts and truncated 0x addresses mapped from native Bittensor accounts. Its follow-up says the wallet now detects TAO balances at those mapped addresses, displays them in the portfolio and offers withdrawal to the native account. This addresses balances that can be difficult to locate after bridges, contract payouts or transfers. The capability is a primary vendor announcement; this review did not connect a wallet, inspect private balances or execute a withdrawal. It is not an announcement that every 0x account is controlled by the same native key.
Shizzy interview calendar: SN71 Leadpoet Thursday, SN116 Carbon Friday
The official YouTube Live tab now lists SN71 Leadpoet for November 5 at 11:00 AM Eastern / 8:00 AM Pacific and the rescheduled SN116 Carbon conversation for November 6 at 1:30 PM Eastern / 10:30 AM Pacific. Both remain upcoming. The site’s featured stream is SN71 Leadpoet; the completed SN65 TPN episode is in replays.
SN23 Trishool publishes HaloGuard 2.0 output-safety release details and paper
At 11:04 AM Eastern on November 4, SN23 Trishool announced HaloGuard 2.0 and linked Apache-2.0 weights and its research paper. Unlike a prompt filter, the 0.8B streaming guard scores response tokens as they arrive, allowing a deployment to stop an unsafe completion during generation. The English Stream model card reports 83.6 average F1 across six benchmarks, or 83.3 on held-out decision-rule splits; these are the authors’ results, not an independent reproduction or a guarantee that harmful output is blocked. October 30’s report already covered the preliminary 83.63 score. The new follow-up is the accessible model, serving configuration and methodology: the paper was submitted November 1 and repository files were uploaded before today’s announcement. The separate multilingual variant’s results should not be attributed to the English checkpoint. No new paying customers or production rollout were verified.
SN31 rec4ll raises miners to 55% and the owner to 25%
Overnight carry-forward, executed November 3 at 11:47:36 PM Eastern: finalized block 9,430,332, extrinsic 22, changed SN31 rec4ll from approximately 18% owner / 41% validators / 41% miners to 25% / 20% / 55%. The first two slots remained zero. Successful execution and exact integer vectors are archived. Historical yield may lag the new allocation; this event does not by itself measure present staker returns. It is the same previously archived transaction, not another change today.
SN36 Weir sets its identity and documents GPU rentals
SN36 changed from an unnamed identity to Weir at finalized block 9,432,987 on November 4, followed by another metadata update at 9,433,014. The owner and registration 9,420,559 stayed unchanged across both transactions. Its public site offers prepaid GPU machines with root SSH; the linked miner/validator repository identifies SN36. The site and code are accessible, but no paid rental, delivered GPU capacity or revenue was independently tested. This is the current registration, not continuation of the earlier SN36 project.
SN44 Score shifts the allocation to 12% owner, 44% miners and 44% validators
Overnight carry-forward, executed November 3 at 8:18:36 PM Eastern: finalized block 9,429,287, extrinsic 8, changed SN44 Score’s five-slot vector from approximately 18% owner / 41% validators / 41% miners to 12% / 44% / 44%. The first two slots remained zero. The permanent evidence retains exact integer vectors and successful execution. This allocation change is not a transaction burn or proof that an owner-wallet lock was executed.
SN51 Lium bounds pod edits when a container will not stop
SN51 Lium merged PR2048 on November 4 at 15:51:11 UTC, or 10:51:11 AM Eastern. A pod edit or reboot could hang when a remote container-stop command never returned. The patch bounds the edit-swap commands and rollback path; if stopping and killing still fail, it restores the container’s name and fails the edit while retaining its volume. The author explicitly does not claim an overall timeout for every create operation, which can include legitimate long image pulls. This is verified merged code, not an independently exercised rental or confirmed fleet-wide upgrade.
SN64 Chutes lists DeepSeek V4.1 Flash at $0.30 input and $1.20 output
At 11:04 AM Eastern on November 4, SN64 Chutes announced DeepSeek V4.1 Flash availability. Its public catalog lists deepseek-ai/DeepSeek-V4.1-Flash-1017 as Hot, with prices per million tokens of $0.30 input, $1.20 output and $0.03 cached input. The same catalog lists V4 Pro at $0.85 input and $2.65 output. Developers can choose between the models through Chutes’ API rather than use the larger model for every call. Listing and advertised pricing were directly checked; authenticated inference, uptime and comparative benchmark scores were not independently tested. This is separate from the still-planned Tencent Hy4 integration.
SN65 TPN publishes the 2.32 GB TPN-007 model
On November 4 at 6:11 AM Eastern, SN65 TPN announced its TPN-007 output. The task was to fit Qwen3.5-4B under a 4.2 GiB ceiling; its public Hugging Face repository contains a 2.32 GB GGUF file. SN65 TPN says it runs in Gratis. The original model URL now redirects to tpn-007-winner. GB and GiB are different units. File availability is verified; retained model quality and local runtime performance were not independently benchmarked.
Watch Shizzy’s completed SN65 TPN interview
The November 3 interview with Mikel from SN65 TPN is now a public replay, not an upcoming stream. The 55:32 episode covers the project’s background, model optimization and Gratis. The YouTube description has 25 chapters and verified project links. The site video library now offers the replay for playback.
SN71 Leadpoet merges recovery for completed paid Arena work
SN71 Leadpoet merged PR1015 on November 4 at 11:32:50 UTC. The patch keeps completed-result delivery running during shutdown, rechecks cancellation and deadlines before provider dispatch, and retains successful sanitized replies when billing persistence fails. Exact same-call retries can reuse the retained result rather than issue another provider request, while unresolved costs remain subject to normal ledger reconciliation. The PR explicitly leaves verifier prompts, judging criteria, historical outcomes and reward rules unchanged. This code merge does not establish a production rollout or resolve the separately reported baseline-verifier mismatch.
SN75 Hippius reports 1 PB stored
Dated November 3 at 2:40 PM Eastern: SN75 Hippius says its distributed storage network holds 1 PB, described as 1,000 TB. That is a team-reported usage milestone. The reviewed announcement does not establish paid storage, unique customer data, replication accounting, utilization economics or monthly revenue. A later “next petabyte” reply describes the next ambition, not a second completed milestone.
SN91 Cascade describes four cross-subnet connections
SN91 Cascade’s November 4 post says SN118 Ditto uses Ephemeris for credit-usage forecasts; SN4 Targon supplied compute credits for architecture testing; SN6 Numinous exchanges knowledge-graph data and forecasts; and SN51 Lium supports GPU rentals for its mixture-of-experts competition. SN6 Numinous also acknowledged the collaboration on November 3. These are reported connections, not four independently measured new production deployments or revenue contracts.
SN92 REAX adds 24-hour notice for reward-rule changes
The SN92 REAX FAQ’s version 3.8, dated November 4 at 11:45 UTC, requires at least 24 hours’ notice for changes to tiers, shares, targets, pools or scoring. Adding an inference model is exempt, but needs a versioned release stating its pool and scoring; removing or replacing a model is not exempt. The page says already-earned amounts are not taken away. Separately, version 3.7 was decided November 2 and announced November 3: the three full-share days start at award approval, or crowning if approval came first, from the next crowning. Current champions get a one-time adjustment without double payment. These are documented rules; this review did not reconstruct individual payouts.
SN117 EveryFrame announces a film-contest treasury from unused miner allocation
Dated November 3, last edited 3:11 PM Eastern: SN117 EveryFrame says unused miner allocation will fund its AI Film Contest rather than be recycled. Serving fees and a 10–50% output-quality bonus are paid to miners first; the treasury receives the remainder of the miner portion, with owner and validator allocations separate. The team identifies treasury UIDs 156/157 and describes community voting and on-chain awards. Its public dashboard currently shows “Treasury data unavailable,” so this review did not verify balances, completed payouts or actual routing. The promotional split graphic is illustrative, not a fixed whole-subnet percentage. This is a team-announced funding policy, not a qualifying transaction burn or measured creator adoption.
SN127 Zils operator arrangement disclosed; Astrid reports a 400 TAO OTC sale
In its November 4 RNS announcement, Astrid Intelligence says Zils, founded by Ken Nicholson, will operate SN127 as an Astrid project. Astrid says it retains the subnet owner key and will recognize its revenues and costs. The issuer also reports selling 200,000 SN127 alpha over the counter at 0.002 TAO each, totaling 400 TAO, with the sold alpha unlocking on an agreed schedule. Its stated remaining position is approximately 3,000 TAO-equivalent, a dated issuer valuation rather than 3,000 liquid TAO. At 1:18 PM Eastern, SN127 Zils described its custom decision-model workflow: customers supply examples, miners compete, validators grade and customers receive an API. The earlier rename and early-access site are already covered above. This follow-up adds the disclosed operator relationship and financing; it does not independently trace settlement, validate a lock schedule, prove customer traffic or establish who bought on the market before the announcement.
Quick briefs & watchlist
6 updatesSmaller developments and early plans. Verification labels distinguish confirmed information from developing claims.
Gamma service-credit proposal remains open
PR3300, opened November 2 and still open at this review, proposes dollar-valued service credits minted by burning subnet alpha. The credits would be non-transferable and non-redeemable, used for a subnet’s service through a hotkey-keyed deposit. Its launch sequence requires an upgrade and enabling steps. It is not a live payment system or an observed transaction burn.
Const argues for protocol-coordinated compute across jurisdictions
Commentary dated November 4 at 8:12 AM Eastern: Const argues that protocols can coordinate decentralized inference, compute clusters and training across jurisdictions, and names SN53 Engy, SN64 Chutes and Instant Inference in that thesis. The post responds to Elon Musk’s remark about data centers in space. This is a view on infrastructure coordination and political concentration, not a new product launch, proof that jurisdictional risk disappears or a partnership announcement.
SN24 ms.computer owner announces a coldkey swap
At finalized block 9,431,616 on November 4, extrinsic 25 emitted ColdkeySwapAnnounced for the verified SN24 ms.computer owner. Adjacent storage still shows the same owner and registration 2,538,424. The event commits to a destination hash; it is not a completed owner transfer, a disclosed buyer or proof of a sale. A later executed swap still requires separate verification.
SN37 Aurelius merges opt-in renewals with a persistent spending cap
PR33 merged into SN37 Aurelius’s main branch on November 4 at 03:19:52 UTC, or November 3 at 10:19:52 PM Eastern. It adds an optional helper to restage a winning entry within an explicit lifetime WT limit and per-evaluation ceiling. The design retains authorization across restarts and uncertain outcomes, stops after a loss or timeout, and waits when evidence or funding is missing. The WT budget is not a TAO-denominated spending claim. The patch does not change coldkey funding, selection or weights. This is merged implementation, not a verified live deployment or uninterrupted-reward guarantee.
SN64 Chutes announces Tencent Hy4 support is coming
On November 4 at 6:16 AM Eastern, SN64 Chutes said it is partnering with Tencent Cloud to bring Hy4 to its existing API under a new model ID and the same key. The post says “coming” and “more soon”; it does not give a launch time, model ID or independently tested endpoint. Treat this as an announced integration, not a verified live deployment. The reviewed quote of Tencent’s model promotion does not independently confirm the commercial arrangement.
SN127 Zils points to small decision models after the Astrid rename
Dated follow-up to the November 3 Astrid → Zils identity change: the current SN127 metadata now links zils.ai and its decision-model repository. The public site describes training small models for routing, flagging and matching tasks, and labels its training workspace early access. Its research and illustrated examples are not proof of general customer results. Sector discovery now follows the current stated product rather than the historical trading-project label. No sale or insider-buying conclusion follows from the rename or chart.
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
Audit trail
Coverage notes
SubConnect estimates $70M in annualized subnet revenue, drawing a response from Barry Silbert. SN4 Targon shows $1.159M in trailing 30-day platform revenue, SN64 Chutes lists DeepSeek V4.1 Flash, and Talisman announces access to mapped-address TAO balances. Earlier verified stories remain below. Afternoon: SN9 IOTA announces open decision-model weights; SN23 Trishool links HaloGuard 2.0 artifacts; Astrid discloses the SN127 Zils operating arrangement and an OTC alpha sale.
Coverage: Nov 3, 2026, 2:07 PM – Nov 4, 2026, 11:14 AM ET
Searched
- 128 active slots reconciled against finalized owner/identity storage and the shared market feed; 49 dated November 4 SubnetRadar briefs read in browser.
- Signed-in Grok discovery, qualitative discussion review and focused omission query; Const Reborn, TensiaFDN, TensiaBot and relevant account activity reviewed.
- Finalized allocation/burn scan, registration/identity/owner lifecycle interval and 128 current-owner holdings; selected primary announcements, model files, repositories, product sites and official YouTube Videos/Live tabs.
Public-data gaps
- Available verified news is published with a source-coverage gap: original X timelines/repositories were not individually exhaustively reviewed for every subnet. The 49 AI-generated Discord digests are discovery, not primary verification; the original Discord link redirected to login. Empty subnet entries do not establish no activity.
- Authenticated TaoStats cross-check is blocked because no local API key is available. No credit balance or indexer completeness was invented. Full private Discord and independently run model/GPU/storage benchmarks are unavailable.
- The promised Tensia dossier and several competition/scoring leads remain unresolved; allegations and unidentified wallets were not promoted to facts. A code merge is not a mainnet deployment, and team-reported adoption figures are not independently audited.
- The 11:02 AM Eastern monitor could not run fresh Grok discovery/omission responses because the signed-in account reached its usage limit. Direct project policies, primary repository changes, account discovery sources and the dated news index were reviewed instead; original-source coverage is not exhaustive.
Browse all 88 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.
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.