Daily digest of the Bittensor ecosystem

中文/EN

Daily edition
Nov 18, 2026

Bittensor Daily A 1T1B product
1T1B ↗
BITTENSOR DAILY

24H Market

01/10

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

THE DAILY BRIEFING

Bittensor Daily — November 18, 2026

Today's clearest news ran through X rather than Discord: Chutes' Alibaba Cloud partnership extends an enterprise push that began with Tencent Cloud earlier this week. The subnet cards carried the heavier substance -- a community investigation into seed-matching patterns on NIOME and Swarm's confirmation that a top-ranked model was cheating -- while #General spent its busiest day of the week on a moderator's extended guidance over insulting emojis and an unresolved quantum-resistant-wallet question, and #Rao quietly corrected a deregistration misconception before falling silent.

Today's highlights

  1. A community investigation on SN55 (NIOME) finds top-ranked submissions matching the specific seeds drawn for their own round, a pattern the team has not addressed.

    (1) One analysis found rank #1 submissions matched round-specific seeds at an 84% rate, ranks #2-4 at 79%, and ranks #5-10 at 16%, across the 25 rounds run since Nov 10. (2) A second community member independently re-checked archived top-10 submissions using the validator's own scoring code and reported the same pattern, proposing a post-scoring audit that re-scores top-10 entries on fresh random seeds. (3) Neither report determined how the seeds might have become available before scoring ended, and the team has so far only announced a storage migration and a monitoring dashboard without addressing the allegation. The question remains open with no official response to the seed-matching reports.

  2. SN124 (Swarm) confirms a top-ranked Swarm Sentinel model was cheating via an exploit, the same day it deploys a new Test Field task.

    (1) The team said the exploit would be fixed and the model banned in the next deployment. (2) The same update, V6.0.1, adds the Swarm Test Field task with its own emission split and a public champion repository, drawing on a real test field in Andorra. (3) Miners separately reported evaluation running slow or stuck, which the team attributed to backend worker and validator queue issues and said it addressed by renting additional machines. The subnet's token fell 2.5% over 24 hours as both the cheating confirmation and the new task landed the same day.

  3. Chutes (SN64) partners with Alibaba Cloud, its second major cloud partnership within a week after Tencent Cloud.

    (1) The deal brings Qwen3.8 Max, GLM-5.3 and DeepSeek V4.1 Flash to developers through Chutes, with a 35% discount on the latter two models running through Nov 23. (2) It follows the Tencent Cloud partnership reported earlier in the week, giving Chutes two of Asia's largest cloud providers as channel partners in quick succession. (3) The announcement landed the same day @taostats rolled out live cross-chain swaps letting users exchange USDC or ETH for TAO or subnet alpha without a bridge or exchange account. No community reaction to either announcement appears in the material.

  4. A #Rao member's claim that age puts the newest 40 subnets in danger of deregistration is corrected: the chain selects by lowest alpha price, not age.

    There will be 40 new subnets in immune period which means the last 40 subnets are in danger in 2 months. drew the correction Chain picks the non-immune subnet with lowest moving alpha price. not age., met with the admission Oops, I thought the oldest subnets like lium and chutes will dereg soon. 😆. (1) This is a plain mechanism correction rather than a debate, clearing up a misconception about deregistration risk. (2) The exchange ran entirely among community members with no moderator or team voice confirming the rule, continuing #Rao's pattern this week of peer answers carrying technical detail. (3) Channel volume fell sharply to 12 messages from 50 on Nov 16, closing out a quiet day with little else beyond this exchange. The original poster accepted the correction and no further question followed.

  5. A #General member asks whether quantum-resistant wallets are fully live on mainnet and how Talisman/Ledger users migrate without losing staking positions, and gets a mechanism answer but no activation confirmation.

    Are quantum-resistant wallets fully activated on Bittensor mainnet yet? drew the explanation You would need to create a new coldkey (wallet) using one of the new key types, and transfer your assets from a current account to the new one., alongside the standing caution Please be warned that if you get any DMs or FRs from people claiming to be support for wallets or wallet applications or quantum resistance (or any kind of support at all, including people pretending to be me or other mods), it's a scam. Block all such requests. (1) This is #General's first substantive engagement with quantum-resistant wallets, days after #Rao walked a member through key-type choices once the TAOApp extension went live earlier this week. (2) None of the asker's three questions -- mainnet activation status, continuity of staking positions, or a migration timeline -- received a direct answer. (3) The moderator's scam warning shows the imposter pattern attaching itself specifically to this new feature. The exchange closed without confirming activation status or a migration timeline.

  6. Other: #General's busiest day of the week was dominated by a moderator's extended guidance on insulting-emoji disputes, alongside routine scam warnings.

    if you insult someone, adding a winking emoji, "j/k", or some other claim that it's a joke is not going to make it okay. (1) #General logged 35 messages, its highest point of the week, driven largely by this drawn-out dispute and the standing scam-imposter script rather than new ecosystem substance. (2) A newcomer's onboarding questions and an unable-to-swap-alpha complaint were both answered with channel redirects and the standard scam script rather than substantive diagnosis. (3) #Announcements logged no messages for the day, continuing its near-silent run through this stretch of the week. Both items closed without further development in the material.

Discord conversations worth following

#General

2 REPORTS

Notable

A member asks whether quantum-resistant wallets are live on mainnet and how existing Talisman/Ledger users can migrate without losing staking positions, and gets a mechanism answer but no activation confirmation or timeline.

Opening ·

A member laid out three concrete questions: Are quantum-resistant wallets fully activated on Bittensor mainnet yet? 2. Will existing Talisman and Ledger users be able to migrate to quantum-resistant accounts without losing their staking positions? 3. Is there any recommended timeline for existing TAO holders to make this transition?

Reply ·

A moderator pointed to existing documentation, Hi, please see this article in the FAQ channel, then explained the mechanism directly: You would need to create a new coldkey (wallet) using one of the new key types, and transfer your assets from a current account to the new one.

Caveat ·

The same moderator closed the exchange with a standing warning rather than further detail: Please be warned that if you get any DMs or FRs from people claiming to be support for wallets or wallet applications or quantum resistance (or any kind of support at all, including people pretending to be me or other mods), it's a scam. Block all such requests.

3 things to read from it

  1. This is #General's first substantive engagement with quantum-resistant wallets, arriving days after Nov 12's #Rao thread walked a member through actual key-type choices once the TAOApp extension went live.
  2. The coldkey-swap mechanic given here matches what #Rao already established on Nov 12, but none of the asker's three original questions -- full mainnet activation status, whether staking positions survive the swap, and a recommended migration timeline -- received a direct answer.
  3. The moderator's closing warning about scammers impersonating 'quantum resistance support' shows the imposter-scam pattern that otherwise fills #General this week attaching itself specifically to this new feature.

In the week’s context

  • Nov 12's #Rao thread first moved quantum-resistant wallets from announcement to usable tooling, with a member walking through which key type to pick once the TAOApp extension went live.
  • Nov 14-15's #Rao custody debate, triggered by SN3 (Teutonic)'s owner-coldkey compromise, discussed wallet security practices more broadly without resolving hardware-wallet or migration specifics -- the same gaps this #General exchange leaves open.

Where it stands

Partially answered: the coldkey-swap mechanism is explained, but mainnet activation status, continuity of staking positions, and any migration timeline remain unaddressed by an official voice.

Other

Other: a prolonged dispute over 'malicious emoji' reactions drew extended moderator guidance, a newbie onboarding question got a channel pointer, a swap-alpha complaint triggered the standing scam script, and routine greetings closed out the day's heaviest #General volume of the week.

Mixed ·

A member asked a moderator what action is taken against people who attach insulting reactions, prompting a lengthy back-and-forth in which the moderator advised If it bothers you that much, react to the post with a hammer emoji. But I really would appreciate it if people could stop calling on the mods to handle schoolyard antics. and later warned the wider channel if you insult someone, adding a winking emoji, "j/k", or some other claim that it's a joke is not going to make it okay. Separately, a newcomer introduced themselves — Hello, newbie here, I'm Ren, I generally work with Voice AI and inference engineering, new to Bittensor, how to get started with the projects? — and was pointed to a channel guide. A member reported Unable to swap alpha, drawing the full standing scam-imposter script rather than a diagnosis. A separate thread noted a subnet's token had launched on Solana: they didn't their Dtao token is now available on Solana (like many other SN tokens).

What to read from it

  1. Volume rose to its highest point of the week at 35 messages, well above Nov 14-16's lows, but the increase was driven almost entirely by a drawn-out moderation dispute over reaction emojis and the day's standard scam-warning repeats rather than new ecosystem substance.

Where it stands

Closed; all items resolved through moderator guidance, channel redirects, or the standing scam-imposter script, with no ecosystem-level question left open.

#Rao

3 REPORTS

Notable

A member's claim that age puts the newest 40 subnets in danger of deregistration gets corrected: the chain selects by lowest alpha price, not age.

Opening ·

A member warned There will be 40 new subnets in immune period which means the last 40 subnets are in danger in 2 months.

Correction ·

Hours later another member corrected the mechanism directly: Chain picks the non-immune subnet with lowest moving alpha price. not age.

Outcome ·

The original poster conceded the point with a joke about which subnets they had assumed were at risk: Oops, I thought the oldest subnets like lium and chutes will dereg soon. 😆

2 things to read from it

  1. This is a plain mechanism correction rather than a debate, but it clears up a misconception that could otherwise shape expectations about which subnets face deregistration risk.
  2. The exchange stayed entirely among community members with no moderator or team voice confirming the corrected rule, consistent with #Rao's pattern this week of peer answers carrying technical detail.

Where it stands

Resolved as a mechanism clarification; the original poster accepted the correction and no further question followed.

Notable

A member questions why subnet owners build products they themselves would not use, reviving the week's running doubt about owner-investor alignment.

Challenge ·

A member asked If subnet owners don't need their own product why do they build it for others? 🙂 I've always thought about users first as a frontend dev. I see many subnets building for investors or some theoretical good but if the product is not something the owners and team are passionate about or don't even need, then I struggle to see who else would want their product

2 things to read from it

  1. The comment stays at the level of a general observation about the ecosystem rather than naming any subnet, and it drew no reply in the channel today.
  2. It lands the same week as Nov 15's unanswered comment that decentralized governance structurally misaligns investors and subnet owners, giving the channel a second unanswered framing of the same underlying worry from a product-design rather than governance angle.

In the week’s context

  • Nov 15's #Rao thread raised a closely related alignment concern -- «There's a problem with alignment between investors and sn owners. A stock is much better aligned» -- which also drew no reply or counter-proposal in the channel.

Where it stands

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

Other

Other: a liquidity-pool question went unanswered, a subnet-cap comment echoed an earlier thread, and the day closed on a riddle joke and an emoji reaction.

Mixed ·

A member asked what's the latest on liquidity pool size plans or discussion with no reply in the material; another member agreed with an earlier point on subnet caps, writing i second this; a member who claimed to have solved a riddle asked Hi [mention] I solved the riddle for getting Gang-signs, can you give me gang-sign? and was told by a moderator Please take a day off to read the article that TAO4Fun has linked to you, thank you.; the day ended with a bare emoji reaction, How'd I do it <:OkayChamp:967797108340064346>.

What to read from it

  1. Channel volume fell sharply to 12 messages today from 50 on Nov 16, with most of the remainder being unanswered or off-topic one-liners rather than substantive follow-up.

Where it stands

Closed through silence or routine redirect; the liquidity-pool question remains unaddressed.

Subnets

5 REPORTS

SN116 24h −6.13%

Carbon

What happened
SN116 (Carbon) experienced repeated grading and queue stalls throughout the day as the team migrated recipe building to GPU machines and grading to new AMD Ryzen hardware, pausing evaluation multiple times. Several miners reported submissions lost or stuck mid-grading during the upgrade. The team said these were infrastructure failures and would be re-run for free without charging a try. Carbon also posted on X that its Motor model matched or beat published literature benchmarks on torque and ripple accuracy on mainnet.
Community
Miners asked about queue speed, grading timing, and whether resubmissions would keep their place in the queue. One community member warned others about a phishing attempt impersonating a support ticket that asked users to connect their wallet.
Official
The team confirmed the grading failures were on its side, said affected tasks would be re-run automatically without charge, and changed the leaderboard to show actual rounded scores instead of rounded-down values. It also clarified that grading allowance is per hotkey and said it will use null consensus rather than decentralized validator consensus to protect exam data.

SN124 24h −2.50%

Swarm

What happened
SN124 (Swarm) deployed Update V6.0.1, adding a new "Swarm Test Field" task with its own emission split and a public champion repository. The team said a top-ranked Swarm Sentinel model was found to be cheating via an exploit and would be fixed and banned in the next deployment. Separately, miners reported that evaluation was running slow or stuck, which the team attributed to backend worker and validator queue issues.
Community
Miners said the new task's emission allocation was too small and asked for it to be raised. There was also back-and-forth among miners over who held the current top ranking, and one miner asked for the flagged model's evaluation to be canceled.
Official
The team said it rented additional machines and was fixing slow-retry handling to address the evaluation slowdown, and later said it would remove unfair zero scores and re-evaluate the affected seeds. The official account also shared the new Andorra test field used for the Test Field task ahead of planned field tests.

SN55 24h +3.49%

NIOME

What happened
A community member published an analysis showing that most top-ranked SN55 (NIOME) submissions in the 25 rounds since Nov 10 produced near-identical stage-3 outcomes only on the specific seeds drawn for their own round, with reported rates of 84% for rank #1, 79% for ranks #2-4, and 16% for ranks #5-10. A second community member independently re-checked archived top-10 submissions using the validator's own scoring code and reported the same pattern, proposing a post-scoring audit that re-scores top-10 entries on about 20 fresh random seeds to flag seed-matched outliers. Neither report determined how the seeds might have become available before scoring ended.
Community
The reports led to extended discussion in the subnet channel, including heated exchanges in which several members accused each other of dishonest mining practices; no resolution was reached in the channel.
Official
The team announced migration of owner validator submission storage to AWS S3 and launched a live Weights & Biases dashboard for miners to monitor validator activity; it has not addressed the seed-matching reports in the channel.

SN105 24h +1.85%

Beam

What happened
SN105 (Beam) excluded large paid 'room' transfers from PRISM scoring after raising self-dealing concerns, following miner reports that a small number of coldkeys held most top incentive UIDs through room-transfer volume. Miners then found that the excluded room bytes still counted toward verified MiB, which sets tier and rank, for roughly 24 hours after the exclusion. The team confirmed the bytes were recovery volume rather than newly counted rooms, and that hourly ':23' qualification rooms never count toward qualified PRISM.
Community
Miners debated at length whether coldkeys controlling most top incentive UIDs through room-transfer volume represented unfair advantage or normal competitive performance, with some presenting detailed board data and others dismissing the complaints as sore-loser reactions.
Official
A team member clarified that hourly ':23' qualification rooms never count toward qualified PRISM, addressing confusion about which room transfers are excluded.

In debate · SN105 room transfer fairness and PRISM exclusion · 10 participants

  • Some miners argued the reward setup favors a few wallet groups that dominate incentive through room volume, disadvantaging other miners.

    “The current reward setup heavily favors a couple of wallet groups, and it leaves every other miner competing at a big disadvantage”
  • Others countered that the system is open to everyone and framed the complaints as an unwillingness to compete rather than unfairness.

    “Why is everyone complaining? You're all complaining about big miners having multiple UIDs, but why not just do the same as them?”
  • A third view welcomed the PRISM exclusion but argued it did not remove the underlying ranking lock caused by capped standard volume versus room volume.

    “Thanks for excluding the paid rooms. That was the right fix. Two points still lock the ranking, and they will not go away”

SN107 24h +0.21%

Minos

What happened
Miners in the SN107 (Minos) channel repeatedly accused an unnamed top miner of cheating through a config-verification system that runs on a private server, and pressed the subnet owner to release validator logs or top configs for independent checking. The owner team said validators had already shared logs with miners who asked, and argued that releasing every round's top config would not stop the complaints. The exchange continued over several days without the top miner responding and without new logs being published.
Community
Some miners insisted that the top miner's silence and the private verification setup pointed to unfair play, while others argued that Bittensor subnets are competitive and miners are not obligated to disclose their methods. A few participants also asked unrelated questions about GATK config tuning during the same period.
Official
The team stated that validators had already shared logs with miners who requested them and that the config-hash comparison against on-chain commitments runs on a private server, publishing only pass/fail results.

In debate · SN107 (Minos) transparency dispute over alleged top-miner cheating · 14 participants

  • Miners demanded the validator logs be released so cheating claims could be checked.

    “Let's release the validator logs, 🙂 If it's implemented, then miners would stop complaining and focus on performance update.”
  • The team said releasing configs would not resolve the dispute and that full top-config disclosure was the only option left.

    “The only way you will be satisfied is if we released every round's top config.”
  • A community member argued that miners staying silent about their methods in a competitive subnet is not evidence of wrongdoing.

    “Asking successful miners to come forward and explain their strategies is something you can do, but if they do not do it, it isn't evidence”

More subnet community coverage every day in the Telegram channel →

Ecosystem debate

7 participants

Discord moderation policy on mocking emojis and minor insults

Community members asked a moderator what action is taken against accounts that attach insulting emojis or mocking comments to others' posts, including an incident raised in the SN107 channel. The moderator responded with a detailed explanation distinguishing harassment from minor insults and clarified how joking indicators like winking emojis are treated.

Positions

  • A community member asked moderators to take action against accounts leaving insulting emojis on others' posts.

    “what actions do moderators take against accounts that attach malicious emojis to others' posts for no reason?”
  • The moderator distinguished harassment, which should be reported to the moderation team, from one-off insults, which users should handle themselves by blocking.

    “If a person is *harassing* you (following you around the server), that's one thing”
  • The moderator stated that adding a winking emoji or a claim of joking does not excuse an insult.

    “if you insult someone, adding a winking emoji, "j/k", or some other claim that it's a joke is **not** going to make it okay”
Official response
A moderator responded directly in the channel, explaining that harassment (being followed around the server) should be reported to the moderation team, while isolated insulting emojis or mocking comments should be addressed by blocking the other account rather than escalating to moderators. The moderator also stated that framing an insult as a joke, including via a winking emoji, does not make it acceptable.
What to watch
Whether further incidents, such as the one raised in the SN107 channel, lead to a more formal written moderation policy rather than case-by-case guidance from moderators.

Voices on X

  • @taodaily_io

    TAO Daily reports that SN64 (Chutes) has partnered with Alibaba Cloud, bringing Qwen3.8 Max, GLM-5.3 and DeepSeek V4.1 Flash to developers with a discount on the latter two models.

  • @trishoolai

    Trishool announced Phase 4 of its roadmap, describing itself as becoming Bittensor's first cyber model lab and starting to earn.

  • @taodaily_io

    TAO Daily reports SN51 (Lium)'s hardware footprint is valued at approximately $50M with monthly billings reaching $2.93M.

  • @conjectures_io

    Conjectures said its miners solved a 56-year-old Erdős and Sárközy problem on reciprocal sums, with the proof verified in Lean.

  • @numinous_ai

    Numinous said its miners forecast an Iranian attack on US targets in Bahrain days before the event, citing a sequence of rising-probability forecasts leading up to the Nov 15 strike.

  • @taostats

    Taostats announced live cross-chain swaps allowing users to exchange USDC or ETH on Ethereum for TAO or subnet alpha tokens without a bridge or exchange account.

  • @taoswap_org

    Taoswap reported that six compute and inference subnets together account for over 45% of Bittensor's subnet emissions, listing Lium at 16.26%, Chutes at 8.73%, Targon Compute at 7.85%, Engyai at 4.77%, Green Compute at 4.23% and Say GM at 4.11%.

  • @yanez__ai

    Yanez AI warned that agents handling payments should not store payment information locally, citing a GitHub attack that stole secrets from at least 1,835 repositories including OpenAI, Anthropic, Cursor and Copilot keys, and said its Pulse product keeps no payment information with the agent.

◆2026-11-18 · End of this edition

Recent editions

Past editions +