
Covenant AI Exits Bittensor | Weekly Bittensor Update
This week, Covenant AI exited Bittensor in the most dramatic departure the ecosystem has ever seen. Sam Dare dumped 37,000 TAO worth of subnet alpha tokens and Const fired back with a new governance proposal called Locked Stake. On the other end of the spectrum, Score (SN44) took first place at Paris Blockchain Week and landed a PwC France alliance. Macrocosmos dropped a research paper achieving 128x activation compression for IOTA (SN9), Chutes (SN64) revealed its fully decentralized team structure, and multiple subnets shipped product updates. Here is everything that happened.
Big Events in the Bittensor Ecosystem
Covenant AI Exits Bittensor After Founder Dumps 37,000 TAO

On April 10, Covenant AI founder Sam Dare announced his team’s departure from the Bittensor network. Within hours, Dare sold approximately 37,000 TAO worth of subnet alpha tokens across Templar (SN3), Basilica (SN39), and Grail (SN81). As a result, TAO crashed more than 15%, dropping from roughly $337 to $284 according to CoinMarketCap data. Trading volume surged by 156%.
Dare accused Jacob Steeves, known as Const, of exercising centralized control over the network. His allegations included suspended emissions, removed moderation rights, and direct economic pressure through token sales. The departure left three major subnets without leadership and erased a significant portion of the March rally that followed the Covenant-72B achievement.
Read the full breakdown of Covenant AI’s exit from Bittensor
Const Responds with Locked Stake Proposal and Subnet Recovery Plan

On April 11, Const published a detailed response on X. He disputed each of Dare’s allegations and described the departure as a personal betrayal. At the same time, he credited the Covenant team for proving that decentralized networks can compete with centralized AI labs.
The most significant part of his response was a new technical proposal called Locked Stake. This mechanism would require subnet owners to lock their tokens for a defined period. Investors would then see in advance if an owner unlocked their position, giving them time to reprice and redirect their stake. Const also confirmed that the code behind all three affected subnets remains fully open source. Community members are already organizing to continue development. Both the Locked Stake proposal and the subnet recovery strategy were discussed during the open Thursday call on the Bittensor Discord.
Read the full breakdown of Const’s response to Covenant AI
Score (SN44) Wins Start In Block 2026 at Paris Blockchain Week

Manako, the commercial product of Score (Subnet 44) on Bittensor, took first place at the Start In Block 2026 startup competition during Paris Blockchain Week. The event took place on April 15 and 16 at the Carrousel du Louvre in Paris, drawing over 10,000 attendees from more than 100 countries. Over 1,000 startups applied for the competition. Only 12 finalists pitched live before a jury that included executives from Bybit, Nethermind, and Ribbit Capital. The total prize pool exceeded $10 million.
Manako also claimed the dedicated Bittensor Track, sponsored by Yuma Group. The win placed a Bittensor-powered project directly in front of institutional decision-makers from firms like Goldman Sachs, JPMorgan Chase, and BNP Paribas. On the same day the conference opened, PwC France and Maghreb announced a strategic alliance with Manako Labs to deploy physical AI at enterprise scale using Score’s decentralized computer vision framework. Combined with an existing partnership with Reading Football Club, Score is building one of the strongest enterprise traction records in the Bittensor ecosystem.
Read the full story on Score’s Paris Blockchain Week victory
Subnet Updates
Score (SN44) and Manako Land PwC France Alliance for Enterprise Vision AI

PwC France and Maghreb announced a formal alliance with Manako Labs to deploy physical AI at enterprise scale. The partnership integrates Manako’s vision AI platform into PwC’s AI and digital advisory practice. Target sectors include retail, logistics, manufacturing, energy, and public infrastructure. The alliance is designed to operate globally through PwC’s network across 136 countries and territories.
Manako is built on top of Score (Subnet 44) on Bittensor. The platform turns existing camera feeds into real-time intelligence systems. Businesses describe what they want to detect in plain language, and Manako delivers automated alerts and actions with no code required. The Score team also reported a 100% conversion rate on 60-day enterprise trials comparing AI performance against human operators. This alliance marks the first time a Big Four professional services firm has formally partnered with a Bittensor subnet project.
Read the full story on the Score and PwC France alliance
Ridges (SN62) Ridgeline Ships Four New Developer Features

Ridgeline, the AI coding agent platform built on Ridges (Subnet 62) on Bittensor, shipped a major product update on April 15. The release adds four features: Rerun, Stop, a collapsible chat view, and per-run commit stats. Together, they give developers real-time control over autonomous agent runs.
Rerun lets developers add context to a previous run and execute it again without starting from scratch. Stop allows killing a running agent mid-task to course-correct immediately. The collapsible chat view organizes all runs tied to a single GitHub issue into a clean threaded interface. Commit stats per run show exactly what changed in each run, with diffs opening in a new tab. The platform still runs inference through Targon (SN4) and Chutes (SN64), composing across multiple Bittensor subnets. The Ridgeline open beta remains free with 10 credits per new account.
Read the full breakdown of the Ridgeline update
Lium (SN51) Upgrades Incentive Mechanism for Better GPU Supply

Lium (SN51) announced changes to its incentive mechanism that give the team more dynamic control over GPU supply on the network. Following the update, providers began supplying nearly every in-demand GPU type, compared to only a few unique GPU models previously. At the same time, the network saw a roughly 20% increase in the total number of GPUs allocated to Lium. The team stated that these changes will help the platform scale up while offering customers a wider variety of constantly available GPUs.
Targon (SN4) Now Powers BrainPlay’s Real-World AI Evaluation Layer

Targon (SN4) announced it is now powering BrainPlay, a platform building the evaluation layer for real-world AI. Built by ShiftLayer, BrainPlay evaluates AI through game environments instead of traditional static benchmarks. Agents must interpret context, plan ahead, adapt to other players, and make decisions under uncertainty.
The platform aims to replace conventional benchmarks that measure static performance with dynamic, interactive evaluations. According to the team, this approach produces benchmarks that are transparent, generate performance signals humans can understand, and translate beyond test environments into real-world applications. BrainPlay runs on Targon’s inference infrastructure on Bittensor.
Macrocosmos Publishes ResBM Paper for 128x Activation Compression on IOTA (SN9)

Macrocosmos AI published a new research paper titled ResBM: Residual Bottleneck Models for Low-Bandwidth Pipeline Parallelism. The paper, available on arXiv, introduces a technique that achieves state-of-the-art 128x activation compression without significant loss in convergence rates and without meaningful memory or compute overhead. The architecture is designed from the ground up for low-bandwidth, distributed communication environments.
ResBM introduces learnable subspace projection layers that reduce the dimensionality of communicated activations while preserving a low-rank identity path through the network. Unlike previous approaches such as Subspace Models, which rely on complex constrained optimization, ResBM supports true end-to-end training where all parameters are updated as ordinary model parameters with standard optimizers. The technology underpins IOTA (SN9), Macrocosmos’ distributed training platform on Bittensor. The paper was authored by Alan Aboudib, Rodrigo Lopez Portillo A., Kalei Brady, and Steffen Cruz.
Chutes (SN64) Reveals Full Team and Decentralized Structure

Chutes (SN64) published a full transparency post revealing its team and operational structure. The platform processes over 100 billion tokens daily and operates with no CEO, no board, and no single point of control. The team consists of ten contributors with funds managed through a smart contract that pays staking rewards to team members. No one person controls the money, the code, or the direction.
The named contributors include Jon Durbin (Backend), Florian (Backend and Research), Kyle Widmann (Backend, built their TEE), Chris (Backend), Vince (Frontend), Timon (Sales and Customer Success), Fezicles (Marketing and Community), Rykorb (Technical Staff), Cxmplex (Backend), and Vonkaiser (Technical Staff). The post signals a deliberate move toward radical transparency at a time when the Bittensor ecosystem is actively debating governance and accountability following the Covenant AI departure.
Synth (SN50) Launches Historical Data in API

Synth (SN50) announced that historical data is now live in the Synth API. The feature is available to users on the Pro and Pro Unlimited plans. Full documentation is available on the Synth API docs page.
BitMind (SN34) Partners with CysecOnline for Deepfake Detection

BitMind (SN34) announced that CysecOnline, a digital forensics firm based in South Africa, is integrating BitMind’s technology into its services. The partnership focuses on detecting deepfakes with the goal of protecting clients from synthetic media, fraud, and misinformation. The integration combines real-time AI verification from BitMind with CysecOnline’s existing forensic capabilities.
TAO Market Update

Price: $236.29 – $279.07
Weekly: -5.81%
Ranking: #32
Market Cap: $2.77B
24h Volume: $385.51M
Top Gainer Subnet: Score (SN44)

Score is a decentralized computer vision framework operating as Subnet 44 on the Bittensor network. Its commercial product, Manako, turns existing camera networks into real-time intelligence systems where businesses describe what they want to detect in plain language. The platform targets sectors including sports analytics, retail, logistics, energy, and public infrastructure. Score recently won Start In Block 2026 at Paris Blockchain Week, secured a strategic alliance with PwC France and Maghreb, and holds an active partnership with Reading Football Club.
Weekly change: +24.41%
Price: $9.332491
Market Cap: $41.54M
Volume (24h): $4.78M
All subnet tokens featured in this update are listed on SimplyTao. The platform supports fiat and crypto purchases with no wallet configuration needed.
Disclaimer: Digital asset prices are subject to high market risk and price volatility. The value of your investment may go down or up, and you may not get back the amount invested. You are solely responsible for your investment decisions and SimplyTao is not liable for any losses you may incur. Past performance is not a reliable predictor of future performance. You should only invest in products you are familiar with and where you understand the risks.
Frequently Asked Questions
On April 10, Covenant AI founder Sam Dare announced his departure from Bittensor and sold approximately 37,000 TAO worth of subnet alpha tokens across Templar (SN3), Basilica (SN39), and Grail (SN81). TAO dropped more than 15% within hours. Const responded on April 11 with a detailed rebuttal of Dare’s allegations and introduced a new proposal called Locked Stake to prevent similar events in the future.
Locked Stake is a proposed protocol-level mechanism where subnet owners lock their tokens for a defined period. This creates a transparent, on-chain signal of long-term commitment. Investors would be able to see in advance if an owner unlocked their position, giving them time to reprice and redirect their stake before any damage occurs.
Manako, the commercial product of Score (Subnet 44) on Bittensor, took first place at the Start In Block 2026 startup competition. The event drew over 1,000 applicants, and the total prize pool exceeded $10 million. Manako also won the dedicated Bittensor Track sponsored by Yuma Group.
ResBM is a new activation compression technique published by Macrocosmos AI that achieves 128x compression without significant loss in convergence rates. The technology underpins IOTA (SN9), Macrocosmos’ distributed training platform on Bittensor. It enables large-scale model training across low-bandwidth, decentralized environments.
PwC France and Maghreb formed a strategic alliance with Manako Labs to deploy physical AI at enterprise scale using Score (SN44) on Bittensor. BitMind (SN34) partnered with South African forensics firm CysecOnline for deepfake detection. Targon (SN4) began powering BrainPlay, a real-world AI evaluation platform built by ShiftLayer.


