Targon x Intel Whitepaper | Weekly Bittensor Update

A Bittensor Hackathon in San Francisco crowned three winners, Const broke down the Bittensor thesis on VirtualBacon, Chutes (SN64) locked in a compute deal with GenLayer, Targon (SN4) co-published a whitepaper with Intel, Hippius (SN75) replaced IPFS with its own storage engine, Score (SN44) launched a private track for high-stakes tasks, and Ridges (SN62) dropped its Q2 roadmap. TAO closed the week up 7.9%.

Big Events in the Bittensor Ecosystem

Bittensor Hackathon Track at SF Crowns Three Winners

Three projects walked away with prizes at the San Francisco Bittensor hackathon track, hosted by Redwood North and backed by Funding Commons’ Intelligence at The Frontier program.

The winning teams each tackled a different corner of the Bittensor ecosystem. Provenonce AI took a prize for Proof of Assurance, a project focused on cryptographic assurance for AI workflows. PCC Protocol impressed judges with its Physical Capability Cloud Protocol, which enables verified coordination for physical-world services. Finally, Insignia by Tech Insignia earned recognition for building decentralized quantitative trading with competitive model evaluation.

The prize pool was stacked. Winners split $3,500 in cash and $3,000 in compute credits from Basilic AI. On top of that, Unsupervised Capital offered up to 1,000 TAO as a discretionary investment. Crucible Labs chipped in 50 TAO toward mainnet registration, and Bitstarter AI provided an intro call for further support.

The event highlights the growing momentum behind Bittensor hackathons as a pipeline for new builders entering the network. With real capital and ecosystem support on the table, these events continue to attract serious talent to the decentralized AI space.

Const Breaks Down the Bittensor Thesis in VirtualBacon Interview

Bittensor co-founder Jacob Steeves (Const) joined VirtualBacon for one of his most revealing interviews to date. The core argument? Money is already an optimization algorithm, and Bittensor applies that logic directly to AI. Each subnet functions as a live market for a specific digital commodity, whether that is inference speed, storage, prediction accuracy, or the value of a gradient update to a machine learning model.

Const went deep on Templar (SN3) and its flagship Covenant 72B model, a 72 billion parameter model fully pre-trained from scratch over the internet. He explained how Templar prices every single gradient contribution by measuring its actual impact on the global model’s loss. The team’s SparseLoCo compression technique has already started spreading beyond Bittensor, with researchers using it to train models across home MacBooks.

The conversation also covered how Dynamic TAO transformed the ecosystem overnight. Before dTAO, a small group of validators decided subnet emissions. Now, each subnet has its own alpha token trading against TAO on a constant-product AMM. The result: every subnet team had to start competing for real demand. Websites, APIs, social media accounts, and podcast appearances followed. Const compared it to a version of Google where every internal team has its own token and must sell itself to the rest of the company.

He closed with a strong take on ownership. Open source gives you access to code, but no stake in the system. Bittensor offers a transparent ownership layer where anyone can participate from the ground floor, long before value gets captured by insiders.

Click here for full Interview Summary.

Subnet Updates

Chutes (SN64) Lands Integration Deal With GenLayer

Chutes (SN64) scored one of its biggest partnerships yet. The Bittensor-powered inference platform is now providing AI compute for GenLayer, a blockchain that uses large language models directly in its consensus layer. Instead of traditional deterministic smart contracts, GenLayer runs what it calls Intelligent Contracts, where validators use different LLMs to evaluate subjective decisions on-chain.

Essentially, the integration solves a real infrastructure problem. GenLayer’s consensus mechanism, called Optimistic Democracy, requires each validator to run a different AI model to prevent any single LLM from dominating outcomes. Building that from scratch would mean every validator managing its own GPU setup. However, with Chutes, they simply connect to the platform and get instant access to over 60 open-source models through a single API, all powered by Bittensor’s decentralized miner network.

On the funding side, GenLayer raised $7.5 million in seed funding led by North Island Ventures, with backing from Arrington Capital and Arthur Hayes’ family office Maelstrom. In addition, the project already has partnerships with ZKsync, Caldera, Radix, and Nansen, and launched Testnet Bradbury in early 2026.

For the Bittensor ecosystem, the deal represents a growing pattern. Subnets are landing real external clients, and as a result, every GenLayer transaction that requires AI reasoning generates inference demand flowing through the network.

We covered this partnership in full details here.

Targon (SN4) Publishes Joint Whitepaper With Intel on Confidential Compute

Manifold Labs, the team behind Targon (SN4), dropped a joint whitepaper with Intel on confidential computing for decentralized AI. The paper outlines how Intel TDX and NVIDIA Confidential Computing can protect AI workloads running on untrusted hardware, a problem that has kept enterprise clients away from decentralized compute.

The core issue is straightforward. When your workload runs on someone else’s machine, the operator can potentially inspect memory, tamper with execution, or extract sensitive data like model weights. Traditional encryption covers data at rest and in transit, but leaves a gap during active computation. The whitepaper closes that gap with hardware-level isolation. Intel TDX creates encrypted virtual machines where even the machine owner cannot see what is happening inside. On the GPU side, NVIDIA’s H100/H200 and B200 series run in protected mode with encrypted CPU-to-GPU communication.

Targon adds its own layer on top. Every node passes remote attestation before it can launch, and the system re-verifies integrity every 72 minutes. Nodes that fail any check get removed from the scheduling pool immediately. Anti-cloning protections prevent operators from copying encrypted disks or replaying attestation from different locations.

This appears to be the first time a Bittensor subnet has co-published a formal security architecture with a major semiconductor manufacturer. For the ecosystem, it signals that Bittensor-powered compute is getting serious about the kind of guarantees healthcare, finance, and defense clients require before moving off centralized cloud.

Click here for more info.

Hippius (SN75) Replaces IPFS With Its Own Storage Engine

Hippius (SN75) announced it has moved on from IPFS entirely. The subnet built its own trustless distributed storage engine called Arion, designed from the ground up for the demands of decentralized storage on Bittensor.

Arion uses a CRUSH map for shard placement, distributing both data and parity shards across SN75 miners. The architecture is built for resilience. According to the team, the network can lose up to a third of its nodes and still reconstruct data perfectly. That level of fault tolerance goes well beyond what most decentralized storage solutions offer today.

The move signals a shift in how Bittensor subnet teams approach core infrastructure. Rather than relying on existing protocols and patching around their limitations, Hippius chose to build a purpose-built solution optimized for its own network.

Score (SN44) Introduces Private Track for High-Stakes Vision Tasks

Score (SN44) is splitting its incentive structure into two lanes. Alongside its existing public track, where open models and shared benchmarks drive fast iteration, the subnet is launching a private track designed for problems that are too complex or too specific for open evaluation.

The public track stays the same. It builds a library of vision primitives that anyone, human or agent, can use. Every improvement compounds, and the open format keeps iteration fast. The private track works differently. Miners run their own evaluation infrastructure, submit weights to Score, and the team deploys them inside Manako. Rewards are higher because the work is harder, with emissions starting at 10% and scaling up to 30%.

The private track targets real-world use cases where clients need closed evaluation, custom infrastructure, and miners willing to go deeper than standard benchmarks. Think problems too sensitive or too specialized to expose in an open competition.

The first private track task launches in early April, and the team is calling it the hardest task ever run on SN44.

Ridges (SN62) Drops Q2 Roadmap Focused on Evaluation and Scaling

Ridges (SN62) published its Q2 2026 roadmap, laying out a packed timeline for its autonomous software engineering subnet. Since launching Ridgeline, the team has been refining agent performance, and the next quarter doubles down on evaluation quality and infrastructure.

March brings four upgrades: synthetic evaluations to improve generalization and reduce overfitting, adaptive screener infrastructure, a new cost scoring metric, and running-time scoring that evaluates agents based on how long they take to complete tasks. In mid-April, the team plans to roll out deterministic evaluation for consistency and fairness, alongside broader inference improvements. By the end of April, Ridges aims to integrate X402 and Handshake payments (SN58), enabling agentic access and streamlined payments across the subnet.

The roadmap shows a clear focus on making evaluations more rigorous and fair while building the infrastructure to scale autonomous coding agents into real production use cases.

TAO Market Update

Price: $280.66 – $374.56
Weekly: +7,9%
Ranking: #28
Market Cap: $3.33B
24h Volume: $430M

Top Gainer Subnet: Targon (SN4)

Targon is a decentralized confidential cloud compute platform built by Manifold Labs, operating as Subnet 4 on the Bittensor network. It provides secure GPU rentals for AI training and deployment, backed by hardware-level protection through Intel TDX, AMD SEV, and NVIDIA Confidential Computing. The platform offers auto-scaling serverless infrastructure with over 1,000 GPUs, sub-50ms latency, and 99% uptime.

Weekly change: +28.65%

Price: $20.405827

Market Cap: $90.34M

Volume (24h): $2.01M

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.

FAQ: