Daily digest of the Bittensor ecosystem

中文/EN

Daily edition
Nov 16, 2026

Bittensor Daily A 1T1B product
1T1B ↗
BITTENSOR DAILY

24H Market

01/10

Gainers
--Loading…--
Losers
--Loading…--

THE DAILY BRIEFING

Bittensor Daily — November 16, 2026

#Rao carried almost all of today's substance: the custody conversation that began Nov 14 around SN3 (Teutonic)'s coldkey compromise extended into concrete personal practices and a sharp disagreement over sharing them in public, while a parallel thread weighed whether a proposed shorting/lending feature is worth the added attack surface. #General stayed at its thinnest point of the week, and the subnet cards carried the day's other concrete news -- Carbon clearing a grading backlog and ms.computer shipping its first release. No new tokenomics or governance resolution emerged beyond echoes of last week's unresolved owner-accountability debate.

Today's highlights

  1. A #Rao member details their own wallet-security setup in public and is told it contradicts the paranoia they claim to practice.

    Spend limits + manual hardware confirmation is what saves ya, I’ve never had any crypto stolen and chutes even our like team payment wallet is a 3/5 multisig that’s not even owner wallet. Pita but better than getting robbed drew the challenge Why are you sharing this in public? Doesn’t match “insanely security paranoid”. Better delete all this., answered with Because my main wallet has $11 in it and They can come for me 🤣. (1) This continues Nov 14's airgap debate into a second day, moving from abstract device design to concrete operational practice -- spend limits, multisig payment wallets, periodic machine wipes. (2) It produced the week's first visible friction over disclosure itself, rather than over the underlying security question. (3) The exchange closed without consensus or any official guidance, and the SN3 (Teutonic) compromise from Nov 11 was referenced only obliquely.

  2. A #Rao member warns a proposed shorting/lending feature risks a Kava-style drain, prompting a converging counter-proposal of capped rollout plus a staged bug bounty.

    let Kava drained pools be (another) warning on how dangerous is to expose a lending protocol at the blockchain level. Just one bug may be needed for total catastrophe and do not implement this. at least not until AI tools have enough training to understand all corner cases drew the reply id rather see it off by default and enabled per subnet with small caps first, met with offer a 100 000 usd bounty for example if someone can drain the pool in test net and slowly raise it up to 1 000 000. (1) This is a substantive risk debate over a protocol feature still in development, with design detail -- capped rollout, staged bounties -- not seen in prior #Rao security exchanges. (2) It frames protocol-level attack surface as the same underlying worry as the day's parallel personal-custody thread. (3) The two participants ended closer together on id still keep the caps tho, testnet bounty plus capped rollout beats either one alone, with no team voice weighing in.

  3. A #Rao member argues decentralized governance structurally misaligns investors and subnet owners, echoing Nov 11's unresolved staker-accountability debate.

    There’s a problem with alignment between investors and sn owners. A stock is much better aligned and If we add a layer of disagreement on top of everything rn in development it would be really bad. (1) This raises the same investor-owner alignment gap Nov 11's fifteen-hour #Rao thread debated between a staker board of directors and a code-enforced alternative, but from a governance-structure angle rather than a proposed oversight mechanism. (2) Unlike that thread, today's comment drew no reply or counter-proposal in the channel. (3) The concern was left standing unanswered at day's end.

  4. SN116 (Carbon) removes the crown-recheck step that had slowed its Motor mainnet lane to roughly one verdict per 45 minutes.

    (1) The team attributed the slowdown to crown rechecks rebuilding models on the grading machine for about two hours each and removed the step so the lane now only grades, following Nov 14's backlog that cancelled roughly 75 queued submissions. (2) It also introduced a one-waiting-submission-per-coldkey limit rolling out in stages and a king-promotion rule requiring a grade margin above 0.0129. (3) A re-registered miner hotkey was refused submission with a direct_provider_registration_unavailable error, and a roughly 9 KiB graph-size limit was flagged as a gap the team acknowledged; the subnet's token rose 1.89% over 24 hours.

  5. Other: #General held its thinnest day of the week, ms.computer shipped its first release as its token fell 17.91%, and taodaily_io's Gamma weekly tally showed Lium minting nearly all of the week's $43,902 in service credits.

    You can ask your questions or use the <#1230665820859007039> channel. (1) #General logged 13 messages, its lowest point on the week's trend, with an unanswered sell-limit-order complaint and routine moderator and EVM-channel redirects filling the day. (2) SN24 (ms.computer) released HOST and RUN features and opened mining to anyone, with CONNECT planned next, even as its token fell 17.91%. (3) @taodaily_io reported SN51 (Lium) accounted for $43,267.36 of the week's $43,902 in Gamma service credits minted Nov 8-14.

Discord conversations worth following

#General

1 REPORT

Other

Other: two separate members sought a moderator privately, a sell-limit order complaint went unanswered, and an EVM-channel question was redirected before the day closed with a greeting

Opening ·

Two unrelated askers wanted a moderator's attention during the day, one plainly (Who are the moderators?) and one with visible frustration (no want to help me ???? i need moderators !!!); both were offered help and said they had followed up by private message, as in i sent you a private message.

Friction ·

A member reported a functional problem with no reply from anyone in the channel: Can't create sell limit order for alpha.

Routing ·

A member asking Is there a channel for evm discussion? was told The EVM channel has been archived., then pointed elsewhere when they asked if #General itself was the right place: You can ask your questions or use the <#1230665820859007039> channel.

Close ·

The day ended on a routine greeting: Good day everyone! Wish everyone a wonderful day and a productive week ahead 🫡

3 things to read from it

  1. Volume fell to 13 messages, the lowest point on the week's trend line, with nothing touching tokenomics, governance or protocol direction.
  2. The unanswered sell-limit-order complaint is a bare functional report with no diagnosis offered, unlike earlier in the week when #General at least supplied a known-issue line for the root-basket freeze.
  3. The same day's #Rao channel carried a distinct EVM question about whether the alpha token exists on EVM and how it works -- #General's EVM traffic today was only about finding the right channel to ask in, not the mechanism itself.

Where it stands

All items closed through private follow-up, redirect, or silence; the sell-limit order issue remains unaddressed in the channel.

#Rao

4 REPORTS

Key

A day-two extension of the airgap-security debate turns into a candid exchange on personal custody habits, including a pushback over discussing security in public

Opening ·

Picking up directly from the prior day's airgap thread, a member restated the design goal: Yeah, so probably one linux, no gui, just to create mulisgn transaction and sign with first address, and other machine to sign and finish with ledger, adding But no matter the complexity of the scheme, if devices are already compromised it just a matter of time before escalation, so its really important to have separate devices to sign.

Expansion ·

A member laid out their own operational posture in a long run of messages: I’d just assume for security posture that your stuff is going to get compromised, maybe you download malware, a discord friend shares it, you pip install a bad GitHub repo, etc many possibilities, concluding Spend limits + manual hardware confirmation is what saves ya, I’ve never had any crypto stolen and describing a team setup where chutes even our like team payment wallet is a 3/5 multisig that’s not even owner wallet. Pita but better than getting robbed.

Pushback ·

Another member challenged the wisdom of sharing this detail publicly: Why are you sharing this in public? Doesn’t match “insanely security paranoid”. Better delete all this., to which the original poster replied Because my main wallet has $11 in it and They can come for me 🤣.

Reply ·

A separate member steered the conversation away from singling anyone out, asking the channel to lets give him a break already should we?, while still agreeing that its important that subnet owners in general understand the seriousness of the situation and start being paranoid about security.

3 things to read from it

  1. This continues Nov 14's airgap debate into a second day, moving from abstract device design to concrete personal practice -- spend limits, multisig payment wallets, and periodic machine wipes -- offered voluntarily by a member rather than prompted by any incident report.
  2. The exchange produced the week's first visible friction over operational-security disclosure itself: a member was told that detailing their own safeguards publicly undercuts the paranoia they claim to practice.
  3. The thread again references the SN3 (Teutonic) compromise from Nov 11 only obliquely, through general references to high-value keyholders being prime targets, without naming the subnet.

In the week’s context

  • Nov 14's #Rao thread debated whether airgapped signing devices actually stop key theft following SN3 (Teutonic)'s Nov 11 coldkey compromise; today's exchange is a direct continuation of that unresolved conversation.
  • Nov 12's quantum-resistant wallet rollout addressed key type but not day-to-day custody discipline; today's thread fills that gap with concrete practices like spend limits and multisig team wallets.

Where it stands

Unresolved as a recommendation: the channel surfaced a range of individual practices and one public pushback on disclosure, but no consensus or official guidance emerged.

Key

A proposal to add a shorting/lending feature to the protocol draws a comparison to a recent DeFi exploit and a counter-proposal for a bug-bounty-gated rollout

Opening ·

A member warned against exposing a lending protocol at the blockchain level, citing an external hack: let Kava drained pools be (another) warning on how dangerous is to expose a lending protocol at the blockchain level. Just one bug may be needed for total catastrophe, and argued do not implement this. at least not until AI tools have enough training to understand all corner cases on wh8ch a blockchain can fail (years).

Reply ·

Another member agreed the attack surface is a real concern but pushed back on blocking the feature entirely: kinda with you on the attack surface part, the lending pr is still fixing HIGH findings round after round, warning it reaches into emission and taoflow math so a lending bug could bleed into the whole subnet not just the loan book, while proposing id rather see it off by default and enabled per subnet with small caps first.

Counter-proposal ·

The original critic countered that caps alone don't neutralize a patient, well-resourced attacker, proposing instead offer a 100 000 usd bounty for example if someone can drain the pool in test net and slowly raise it up to 1 000 000, and if 6 months nobody claim bounty then roll implement slowly, and separately urged priority should be to implement a bug bounty program right now for the features we have with good bounties before increrasing risk surface.

Outcome ·

The capped-rollout proponent conceded the patience point but held their position: fair, a patient attacker just sits and waits for the big pools so caps alone dont save you and id still keep the caps tho, testnet bounty plus capped rollout beats either one alone, adding a proper bounty for whats already live, feels overdue honestly.

3 things to read from it

  1. This is a substantive risk-management debate over a protocol feature still in development, moving from a blanket objection to a converging position that combines capped rollout with a staged bug bounty -- a level of design detail not seen in prior #Rao exchanges on security.
  2. The debate extends the channel's live concern with key and infrastructure security beyond personal custody (the parallel thread in the channel today) into protocol-level attack surface, treating both as facets of the same underlying worry.
  3. No team or official voice weighed in on either the feature's design or the bounty proposal; the exchange ran entirely between two community members converging toward a shared recommendation.

Where it stands

Open: the two participants ended closer together on a capped-rollout-plus-bounty approach, but nothing here reflects an official decision or commitment from any team.

Notable

A member shares an outside warning that decentralized governance structurally misaligns investors and subnet owners

Opening ·

A member linked an external post and reacted: This scared me about how tao will eventually get more decentralized and If it happens soon we will die, mechanisms are not good enough yet.

Expansion ·

They argued the core issue is structural: There’s a problem with alignment between investors and sn owners. A stock is much better aligned and warned If we add a layer of disagreement on top of everything rn in development it would be really bad.

2 things to read from it

  1. This echoes Nov 11's fifteen-hour #Rao debate over whether subnet owners owe stakers formal accountability, raising the same investor-owner alignment gap but from a governance-structure angle rather than a proposed oversight mechanism.
  2. Unlike Nov 11's thread, today's comment drew no reply or counter-proposal in the channel, leaving the concern unanswered rather than debated.

In the week’s context

  • Nov 11's #Rao thread split between a staker board-of-directors proposal and a code-enforced transparency alternative over the same investor-owner alignment question, ending without consensus; today's comment raises the identical concern without engaging that prior debate.

Where it stands

Unanswered: no community member or official voice responded to the alignment concern in the channel today.

Other

Other: alpha-on-EVM questions went unresolved, and the day closed with jokes and emoji reactions

Support ·

A member asked does alpha token exist on evm? if so, how does it work? and was told every evm address has a mirrored coldkey address as well, but a follow-up question on the actual association command -- someone please help me understand how `btcli evm associate` command works? and coz I see error saying error: no EVM key 'default' in wallet -- drew only an emoji reaction with no working answer.

Other ·

The rest of the day's traffic was light-hearted: one member joked In these dark times, I would not mind a motivational message from our Great Leader, Kim Jong Const, and another wrote We need more Exploit the age of AI Security is upon us.

What to read from it

  1. The unresolved `btcli evm associate` error surfaced the same day a #General user independently asked whether that channel was a suitable place for EVM questions -- the two channels left the same underlying EVM-tooling friction unaddressed without cross-referencing each other.

Where it stands

Open on the EVM association error; closed on the lighter remainder with no further discussion.

Subnets

5 REPORTS

SN116 24h +1.89%

Carbon

What happened
SN116 (Carbon) shipped a Motor 1.8 update that removes crown rechecks and lets the official grade decide the crown directly: a contender becomes king if its official grade exceeds the current king's by more than 0.0129, and an early-stop now cuts models certain to time out. Before the update, the Motor mainnet lane had slowed to roughly one verdict per 45 minutes because crown rechecks were rebuilding models on the grading machine. The team also introduced a one-waiting-submission-per-coldkey limit rolling out in stages, and a re-registered miner hotkey was refused submission with a direct_provider_registration_unavailable error. Separately, miners hit a roughly 9 KiB graph-size limit when staging Level 4 documents for Motor 1.8/Cooling 1.8.
Community
Miners asked why grading had slowed to about one verdict per 45 minutes or less and sought an ETA for faster grading. After the update, miners questioned how the new 0.0129 margin changed the score needed to dethrone the king, with some confused about why a lower score had become the new king under the old gap formula.
Official
The team said crown rechecks, which rebuilt models on the grading machine for about two hours each, were the main cause of the slowdown and have now been removed so the lane only grades. It also acknowledged the roughly 9 KiB graph-document limit as a gap on its side and said a fix for the re-registration submission error was coming.

In debate · SN116 Motor 1.8 grading slowdown and new crown margin · 11 participants

  • A miner reported the grading lane had nearly stalled and asked whether a crown recheck was holding it up.

    “since the queue-place update went live (~23:42 UTC) the door has given 4 verdicts in ~2.5 h (about one per 45 min)”
  • The team attributed the slowdown to crown rechecks repeatedly rebuilding models on the grading machine, now removed.

    “Each crown recheck rebuilt a model several times on the same machine that grades, holding the queue for about 2 hours each.”
  • A miner questioned why a score below the previously assumed gap threshold was still accepted as the new king.

    “But why the 0.66 was approved as new king? There is no gap now?”

SN24 24h −17.91%

ms.computer

What happened
SN24 (ms.computer) released its first version, going live with two features: HOST, which lets users host websites or deploy apps, and RUN, which provides sandboxes for AI agents. The team said a third feature, CONNECT, for compute-sharing, is planned next. Mining on the subnet is now open to anyone, after a period of testing with trusted miners.
Community
Community members welcomed the launch with congratulatory messages. Some miners asked technical questions about registration, rate limits, and Mac support; the team directed detailed mining support requests to the Ms. Computer Discord.
Official
The team announced the first release is live on SN24, described the HOST and RUN features, and said CONNECT is coming next. It confirmed mining is open to anyone and that miner support questions should go to the Ms. Computer Discord.

SN44 24h +0.18%

Score

What happened
Miners on SN44 (Score) repeatedly flagged ground-truth annotation issues across the Petrol, Vehicle, Person, PPE, Beverage and Dirtiness elements, prompting the team to review and remove several disputed blocks. A separate dispute arose over whether zero-balance (Chutes) failures should be excluded from the average score the same way warm-up issues are; the team confirmed such failures still score zero and are excluded from averages. The Petrol, Vehicle and PPE elements intermittently stopped processing new rounds during the day, requiring team checks. The subnet's open computer-vision model competition was also reported to power Manako's Vision Agents, which has monitored 124 unmanned Avia Xpress fuel stations in France since October.
Community
Multiple miners across different task elements submitted block-specific complaints about ground-truth labeling, asking the team to review and correct what they viewed as annotation errors. A separate group of miners argued that excluding zero-balance failures from average scores while counting them as zero was inconsistent with how warm-up issues are treated.
Official
The team removed several disputed blocks in the Person, Vehicle, Petrol Station and Room Dirtiness elements after review, while confirming other challenges were correctly annotated, and clarified that zero-balance failures count as zero but remain excluded from average score calculations.

In debate · Ground-truth annotation accuracy on SN44 · 12 participants

  • A miner argued a ground-truth bounding box in the Petrol element was too small, causing a correct detection to be marked incorrect.

    “This miner detected all pumps and hoses correctly, but only the pump on the left annotated with a large box.”
  • Another miner argued ground-truth labels in a Beverage block misclassified items, counting a false positive as true and a correct detection as false.

    “the red circled item is an exact FP. but the yellow circled item is not a can. but that was treated as a TP.”
  • The team pushed back on some complaints, describing them as edge cases consistent with the scoring methodology rather than annotation errors.

    “it’s just an edge case. Not sure what you mean though. Is this a GT issue, or a problem with someone else’s model?”

SN120 24h +0.21%

Affine

What happened
SN120 (Affine) v2 moved through multiple rounds, from r000012 to r000021, with a new king emerging each round -- UID 878, UID 15, UID 979, UID 56, and UID 64 -- each posting incremental score deltas over full 2,000-task evaluations. A miner reported taking a 9B model from 60 to 90 on AIME using ten LoRA updates and 30 million tokens over 33 hours.
Community
Miners repeatedly asked why the eval queue no longer follows registration-block order and requested a real-time dashboard showing UID evaluation order, as was available in v1. Miners also asked for the registration fee to be lowered and for upload-link wait times to be shortened for newly registered UIDs.

SN55 24h +0.67%

NIOME

What happened
A miner shared a local stage1-5 analysis of public submission files, stating that top-scoring JSON files score well only on the current round's specific seeds and do not generalize to other seeds, which the miner said points away from a general ML or biology-based strategy. Separately, a community member claimed subnets are being decommissioned at a rate of one every two days, raising concern among SN55 (NIOME) participants about the subnet's status.
Community
Participants debated whether the competition is currently fair, with one member asking directly if it now is. Others questioned the intent behind the decommissioning claim, and some reassured that the subnet is not affected by such comments, with at least one miner saying they intend to keep mining.

More subnet community coverage every day in the Telegram channel →

Ecosystem debate

3 participants

Securing subnet owner and operator wallets

Following past wallet compromises in the ecosystem, community members discussed how subnet owners and operators should protect their keys, covering device separation, multisig setups, hardware wallets and spend limits. A team member from SN64 (Chutes) noted that the team's own payment wallet uses a multisig arrangement, though this is separate from subnet owner keys. No subnet-wide standard for key security was referenced in the discussion.

Positions

  • Some participants argue that layered setups with separate signing devices are necessary because a single compromised device can eventually lead to full key exposure regardless of scheme complexity.

    “if devices are already compromised it just a matter of time before escalation, so its really important to have separate devices to sign”
  • Others favor an assume-breach mindset, treating any device as potentially compromised through malware, bad downloads or shared files from contacts.

    “I’d just assume for security posture that your stuff is going to get compromised, maybe you download malware”
  • A practical safeguard favored by some is combining spend limits with manual hardware confirmation, which one operator said has prevented any crypto theft for them.

    “Spend limits + manual hardware confirmation is what saves ya, I’ve never had any crypto stolen”
Official response
There has been no official response from Opentensor or any subnet team on standardizing wallet security practices for subnet owners or operators.
What to watch
Whether subnet owners adopt formal practices such as multisig or spend limits voluntarily, or whether further wallet compromises prompt a broader push for standardized security requirements.

Voices on X

  • @opentensor

    Opentensor's weekly ecosystem recap highlights Lium's nearly $35M annualized GPU rental run rate alongside updates from Hippius, Almanac, Beam, Score and OpenRoboto.

  • @taodaily_io

    TAO Daily reports that SN105 (Beam) went live with a Chainlink CRE integration, letting b1m_ai record data transfer reports as attested receipts on Base mainnet.

  • @2xnmore

    A community member describes SN44 (Score) as an open computer-vision competition whose winning models power Manako, a camera-based system monitoring 124 fuel stations in France.

  • @TatsuEcosystem

    TatsuEcosystem's weekly brief announces TatsuOS is live, letting users buy plans and AI credits with a listed token, USDC or USDT, credited automatically on-chain.

  • @CryptoZPunisher

    A community member discusses SN24 (Ms. Computer), noting its November 15, 2026 launch announcement aligns with prior speculation about the project's direction.

  • @chainlink

    Chainlink's adoption update lists 10 recent integrations across 5 services and 7 chains, including users such as b1m_ai among others.

  • @taodaily_io

    TAO Daily's Gamma weekly report shows $43,902 in service credits minted by burning subnet alpha from Nov 8-14, led by SN51 (Lium) at $43,267.36.

  • @KakashiTao

    A community member welcomes SN24 (Ms. Computer) as the third new subnet launched this year, alongside SN28 and SN53, expressing bullishness.

◆2026-11-16 · End of this edition

Recent editions

Past editions +