'It felt like a Bitcoin conference in 2013.' - @willobrien, co-founder of @BitGo and CEO of @provenonce_ai Give your feedback so we can build it even better next year: forms.gle/vsJYBqUMAT2jjVAJA

At the booths, on stage and in conversation. More moments with the teams who supported Exploit Summit.

Thank you to the teams who backed Exploit Summit in Montreal. Your support helped bring the people building Bittensor into the same room, with space to show their work and debate where the network goes next. Our sponsors: Affine @affine_io Chutes @chutes_ai Opentensor Foundation @opentensor Proof of Talk @proofoftalk TaoStats @taostats Manifold @manifoldlabs Lium @lium_io KubeTEE @KubeTEEAI Score @webuildscore BigTensor @bigtensorai ChronoLLM @chronollm Dendrite @DendriteHQ Talisman @wearetalisman Yuma @YumaGroup 404gen @404gen_ Bitcast @Bitcast_network BitMind @bitmind Crucible Labs @CrucibleLabs Gradients @gradients_ai Green Compute @green_compute_ OpenRoboto @openroboto Leadpoet @LeadpoetAI Mentat Minds @mentatminds Yanez AI @yanez__ai Thank you also to our media partners: @CryptoAmb · @BitcoinNews · @Cointelegraph · @taotimesdotai · @taodaily_io And to the teams who supported the event: @bitstarterAI · @HackConcordia · McGill AI Lab · @SubnetsForGood_ · @btlabs_ai · @vidaio_ To everyone who showed up, shared their work and made time for a conversation: thank you for making Exploit Summit a place to meet the people behind Bittensor.

That's a wrap! Thank you to every builder, researcher, miner, validator, investor, speaker, sponsor and curious newcomer who came to Montreal and made Exploit what it was - and everyone who tuned in from around the world. Bittensor lives online. But sometimes, you need to gather. To put faces to usernames. Challenge ideas in real time. Make new connections IRL. Exploit was much bigger than the room and we can't wait for the next one🧡

Day two at Exploit Summit, Montreal. Carbon won Pitchtensor. Gamma got a deeper technical discussion. Minos demonstrated an agent-powered research app. Across both stages, builders showed their work and debated what Bittensor needs next. Here’s the recap. OTF STAGE • @const_reborn and @KeithSingery explored the Gamma proposal: on-chain service credits, predictable compute costs, and the trade-offs when subnet owners spend emissions on other subnets’ services. • @carbonphysicsai won Pitchtensor, reaching 500 TAO. @macrozack announced a round lasting 19 minutes and 58 seconds, with Carbon and @fugal_ai pitching through @bitstarterAI. Congratulations to @RyanBequette and the Carbon team. • @centrum_blue announced limited @theminos_ai Studio access for approved academic researchers, combining synthetic genomic datasets with miner-tested configurations. He also previewed Minos’ second incentive mechanism: an agent research loop on a Mac, with an iPhone companion to pause the miner remotely. • Ben James showcased @404gen_ game jam winners and outlined the next step: helping creators monetize their games, with revenue shared with SN17. • @WSquires and @mogmachine debated how emissions should support growing businesses and research, including useful outputs, accountability and timelines. • @tadlouni examined what subnet participants actually control: keys, operations, emissions and decision-making. • @yubrew of @bitsecai explained findings in @heydittoai’s incentive mechanism and why seed timing matters. He said the reported issues had been fixed. • In the final panel, @jon_durbin of @chutes_ai walked @GSchvey through adopting conviction lock: a safer result, alongside the development work required to accommodate protocol changes. RAO STAGE • @LindsMikeStone of @YumaGroup, Kelly Woodward of @CrucibleLabs and @capradavis joined @macrozack of @bitstarterAI to discuss how incubators select and support founders, and why a team’s ability to adapt matters beyond its original idea. • @josercaldera demonstrated Yanez Gate and Pulse: an agent’s spending request blocked by its allowance, then authorized after the limit changed. • Will O’Brien showed Provenonce’s first Bittensor testnet cycle and its observatory: one validator and two miners, all operated by the team. Test TAO, not mainnet. • Youssef El Moujahid of @YumaGroup presented ASMR, a research proposal for adjusting subnet liquidity using slippage benchmarks and a volatility limit. • @quack_builder showed the device behind @openroboto Shift and described how data sales could support robotics research, with data returning to the open pool after a buyer’s exclusivity period. • @isabella618033 showed how @ConnitoAI turns training requests into reusable experts, with requests open for review and a model marketplace waitlist. • @xavi3rlu announced a partnership between @pareton_ai and @engyai to optimize inference for lower latency and higher throughput. • @RomAI__ and @0xgrizzz of @reliquary_ai invited subnet teams into early access for post-training on customer tasks. • @Nic_B1m introduced @b1m_ai Studio, SDK and CLI, alongside an Action Registry where developers can publish extensions and charge for their use. • @Gareth_Vidaio described @vidaio_ powering the event’s multilingual stream and the team’s streaming and object-removal launches that week. • @peytonspencer and @omarzanji presented @heydittoai V2’s shared context and portable memory, plus a planned Router track rewarding lower-cost coding workflows without sacrificing quality. • Josh Riddett of @green_compute_ announced another $4 million data centre planned for the end of October. Thank you to the speakers, builders, miners, validators, researchers and community members who made these two days possible. For the demos, the difficult questions, the pitches and the conversations that carried on beyond the stage. From Montreal, thank you for being part of Exploit Summit. We’ll share any further updates in the posts that follow.

A Mac app for miners. An iPhone companion to control it remotely. “Let your agent do the work for you.” @centrum_blue previews @theminos_ai’s second incentive mechanism: AI agents working on genomic research tasks for Bittensor’s SN107. Watch the on-stage demo run a research loop, open its logs and pause the miner from a phone.

“The acceleration, to an extent, makes us pause and decelerate.” @jon_durbin of @chutes_ai walks @GSchvey through the work behind Chutes’ move to conviction lock. He says the result is safer, but explains how protocol changes can interrupt product development. A builder’s perspective from Exploit Summit’s closing panel.

Another $4 million data centre, planned for the end of October. At Exploit Summit, Josh Riddett of @green_compute_ announces the next deployment for Bittensor’s SN110, following the $3 million facility he says recently went online.

“So trust is the next bottleneck here in physics AI.” Fast predictions only matter if engineers can rely on them. At Pitchtensor, @RyanBequette explains the problem @carbonphysicsai is tackling: proving where a model works, under which conditions, and where its limits are.

The livestream you're watching is powered by @vidaio_. @Gareth_Vidaio confirms the team is streaming Exploit Summit in multiple languages and has launched two products this week: streaming and object removal. Bittensor’s SN85, in use at the event itself.

Build an extension. Let other teams use it in their data transfers. @Nic_B1m announces @b1m_ai’s Studio, SDK and CLI, alongside an Action Registry where developers can publish extensions and charge for their use. New tools for building on Bittensor’s SN105, unveiled at Exploit Summit.

Off stage, still in the conversation. @const_reborn spending time with the Bittensor community, taking questions and hearing people out. Another side of Exploit Summit. And there’s more to come.

Your AI tools shouldn't need the same briefing every time you switch between them. @peytonspencer and @omarzanji explain how @heydittoai V2 connects chats and agents through shared context, with portable memory and the option to run Ditto on your own infrastructure. From Bittensor’s SN118 at Exploit Summit.

A new challenge for SN118 miners: make coding tasks cheaper without sacrificing quality. @peytonspencer of @heydittoai outlines a planned Router track, using real Claude Code and Codex tasks to evaluate cost savings and memory context. Submissions that hurt quality would be penalized. With cofounder @omarzanji at Exploit Summit.