Max

5.3K posts

Max banner
Max

Max

@MaxScore

co-founder and ceo @manakoai | core-contributor @webuildscore subnet 44 & @sire_agent ⎸ prev: @crunchdao ⎸ opinions are my own

Paris, France शामिल हुए Ağustos 2017
2.5K फ़ॉलोइंग5K फ़ॉलोवर्स
Max
Max@MaxScore·
“extra, extra, this week’s TAO times is out”
TAO Times ⚡️@taotimesdotai

TAO Times #49 | Resilience, resistance, and Breakout 👇 Fresh out of Breakout in SF - and the inbound interest in Bittensor is unlike anything since dTAO launched. This week: TargonOS drops, @Grayscale moves closer to putting TAO on NYSE, and @WallStreetBets gets TAO pilled ⚡️

English
0
1
32
1.3K
Tseu Tseu - τao
Tseu Tseu - τao@tseutseutao·
$TAO #SN105 - @b1m_ai is starting to make serious waves and about to doxx the full team. And look who's passing by their Discord channel... Hello @MaxScore 😄 More details about Beam below. Read on to the end to understand just how much of an asset this subnet can be for the ecosystem and why our beloved Max is undoubtedly already showing interest here. -- In short, BEAM is a decentralized bandwidth coordination layer. While other subnets focus on providing storage (#SN75) or providing inference (#SN64), BEAM focuses on moving data across the globe at peak efficiency. 1. The Core Innovation: Proof-of-Bandwidth Unlike traditional ISPs or cloud providers where you pay for "theoretical" speeds, BEAM uses Proof-of-Bandwidth. Verifiable Data: Every transfer generates cryptographic proof. Performance-Based Rewards: Miners are rewarded based on actual throughput, latency, and reliability—not just for being "online." Eliminating Cheating: You can't fake speed; the validators check the math on every chunk of data moved. 2. The Architecture: Miner roles in the Network Orchestrators (The Dispatchers): They act as the "brains." They split data into chunks, select the fastest routes, and balance the load across the network. Workers (The Data Movers): These are the nodes actually streaming the data. They compete to provide the lowest latency and highest bandwidth to earn rewards. 3. The "Why": Why does Bittensor need BEAM? Think of Bittensor as a Global AI Supercomputer. @hippius_subnet #SN75 is the Hard Drive. @lium_io #SN51 is the GPU/CPU. @b1m_ai #SN105 is the Bus. Without BEAM, moving massive AI datasets or model weights between decentralized GPUs would be slow, expensive, and centralized. BEAM ensures the "pipes" of the decentralized web are just as fast as the "brains." As AI agents become more autonomous, they need to talk to each other and move data across different clouds (AWS, Google, Private) without friction. BEAM provides the universal, verifiable "highway" that makes this cross-cloud movement possible without relying on a single corporate middleman, thus unifying all the subnets' communications.
Tseu Tseu - τao tweet media
Beam - Subnet 105@b1m_ai

We have proven that BEAM and its distributed bandwidth network can operate at the same level as traditional transfer providers while remaining fully decentralized. 500MB transferred in 21–29 seconds is still early, but the progress is real.

English
7
9
56
7.8K
Max
Max@MaxScore·
big brains compound
SIRE@sire_agent

Introducing SIRE Core Contributor: Jordan Moore Jordan is a quant researcher with a PhD in Mathematics whose career has been built on one principle: edge comes from better data, better market understanding, and better execution, not from any single model. His background spans modelling, data science, and building the production pipelines that connect the two, with experience across gaming, analytics, and web3. Before SIRE, Jordan worked on forecasting systems, simulation models, and end-to-end data pipelines, consistently in roles where the job was to turn messy data into usable signals. That combination of statistical rigour and engineering pragmatism is what makes his contribution distinctive: he doesn't just build models, he builds the systems that let models run reliably. Focus on Systematic Research As a Core Contributor, Jordan's focus is on building a more systematic and scalable quant foundation for SIRE, stronger research workflows, cleaner data infrastructure, and a clear path from idea to testing to live deployment. In a protocol where intelligence drives capital, the quality of that research layer directly determines performance. A word from Jordan: "What excites me most about SIRE is that this is still a young space, which means there is real room to build an edge rather than just chase one. Prediction markets bring together sport, data, market behaviour, and execution in a way that is genuinely interesting. That makes it a great environment for rigorous research, but also for building things that can have an immediate real-world impact." Jordan brings the qualities SIRE needs at this stage: mathematical depth, production discipline, and a bias toward systems that compound. His work is helping ensure that as SIRE scales, its intelligence layer scales with it.

English
1
1
30
1.3K
Max
Max@MaxScore·
when @Nic_B1m pitched me his subnet idea i knew immediately he was the right person to tackle this decentralised bandwidth problem first milestone was simple and super clear: → move 500MB in under 30 sec, consistently, with just 20–30 workers. ✅ @b1m_ai miners hit this yesterday. it's done. on par with AWS and Google. and they just started 2–3 weeks ago, let that sink in...
Beam - Subnet 105@b1m_ai

We have proven that BEAM and its distributed bandwidth network can operate at the same level as traditional transfer providers while remaining fully decentralized. 500MB transferred in 21–29 seconds is still early, but the progress is real.

English
2
12
69
3.3K
Tansu Yegen
Tansu Yegen@TansuYegen·
A new AI technology called LitterCam in England can see trash thrown from cars and connect it to license plates to help authorities give fines 🚗🗑
English
179
486
2.5K
356.2K
Max
Max@MaxScore·
@Rapido_ai still room for improvement, COME ON YOU MINERS ;)
English
1
0
4
159
Rapido
Rapido@Rapido_ai·
@MaxScore So bad it doesn't target this semi-truck.
Rapido tweet media
English
1
0
2
174
Max
Max@MaxScore·
baseline: 40.4% target: 90.0% best miner so far: 86.9% the detect_vehicle skill is looking guuuud
Max tweet mediaMax tweet media
English
2
9
79
2K
Max रीट्वीट किया
Beff (e/acc)
Beff (e/acc)@beffjezos·
The ultimate moats are grit and agency
English
50
131
1.3K
54.4K
Max
Max@MaxScore·
@MutantApeJack @officialpbuster yeah and the biggest hack so far is still to actually just take the time to engage with people and care about their messages
English
0
0
2
55
Max
Max@MaxScore·
grew our community back then from a couple thousand to where it is today (11k+) with one repeated LinkedIn campaign if you’re a subnet owner looking to attract more miners, here’s the exact playbook: step 1 - scrape > use @officialpbuster to pull profiles that liked or commented on huggingface/kaggle/autoML…posts in your field (computer vision for us) step 2 - filter > keep only kaggle masters, grandmasters, or anyone who mentions kaggle performance on their profile step 3 - outreach via @lemlist > send a LinkedIn connection request with no note, then set up a two-message sequence: message 1 (day 1) - saw your kaggle work, genuinely impressive - we run a similar competition/this our biggest challenge rn - top earners are making up to [x] - here’s the starter pack: [link] message 2 (day 3): - just making sure this didn’t get buried - no deadline, no pressure - but here’s the latest winning model, I think you could beat it - let me know if you want in - we have a discord channel with already 100s of data scientists in your field discussing the the right approach then let your leaderboard do its thing rince and repeat it is that simple
0xSammy@0xSammy

Also, I Intend to catalyze thousands of fresh ML/AI engineers into the Bittensor ecosystem with Crunch: x.com/0xSammy/status…

English
3
6
63
5.1K
Samuel Cardillo
Samuel Cardillo@CardilloSamuel·
decided to rent 8x H100 to train qwen3-coder-next with an opus4.6 reasoning dataset (the same used for the qwen3.5 distillation). let see how it goes!
Samuel Cardillo tweet media
English
18
4
143
10.2K
Max
Max@MaxScore·
@wouterhar and then you’ll have to grind even more
English
0
0
1
165
Wouter | SN18 Zeus ⚡️
We'll keep grinding until the first major b2b pilot converts into a paying customer. $TAO
Mariuszek@sobczak_mariusz

If you remember anything from our AMA let it be this : @zeussubnet is not a weather app. It is a B2B weather intelligence product aimed at energy traders and grid operators. I walked into the $TAO @zeussubnet AMA skeptical and walked out genuinely impressed. Not because @wouterhar sold me. He did not try to. He answered questions directly, acknowledged the problems, and described a business model I had not fully understood before. That combination of honesty, specificity, and a coherent path to revenue is rare in this ecosystem. @zeussubnet has been misunderstood, partly because it has not always communicated clearly, and partly because “weather forecasting subnet” sounds niche if you do not understand the market it is actually serving. @zeussubnet is not a weather app. It is a B2B weather intelligence product aimed at energy traders and grid operators. That matters because energy trading, especially in renewables, is fundamentally a weather problem. Wind, cloud cover, and local weather conditions directly affect output, pricing, and hedging decisions. Companies already pay serious money for weather data because the signal matters. @wouterhar explained the differentiation clearly: @zeussubnet incentivizes miners against the exact benchmarks clients care about in an ongoing competition. That means specific variables, specific regions, and specific time horizons. Traditional weather vendors build one model and sell the same output to everyone. @zeussubnet does the opposite. Instead of one static model, it runs continuous optimization around what a given client actually needs. One of the most important updates from the AMA was the v2 launch. Forecast horizon expanded from 7 days to 14 days, and the incentive structure was cleaned up by moving to winner-takes-all per variable while removing the latency incentive. The 14 day extension matters because it directly matched the requirements of their first pilot client. If the pilot needs a 14 day horizon and you only have 7, you do not have a real pilot. The pilot itself is the most important near-term catalyst. It is a real client using @zeussubnet for weather forecasting tied to energy trading decisions. That does not mean revenue is confirmed. It does mean this is no longer just a story. @wouterhar also said real users care about exact locations at lat/lon precision, exact variables and metrics, and success measures that go beyond standard RMSE. Traders care about whether the forecast improves a decision at the exact coordinates of a real asset. The moat here is not secret technology. The data is public and the code is open source. The edge is the feedback loop: continuous client-specific benchmark optimization wired directly into miner incentives. That is structurally hard to replicate without rebuilding the entire incentive architecture. The go-to-market is still early. @zeussubnet appears to have more than one pilot underway, with at least one live pilot tied to an energy trading client. The product thesis makes sense. Now it has to prove the sales motion. @zeussubnet is not following a standard venture-backed path. It is being built through Bittensor emissions, using crypto-native funding rather than institutional capital. The biggest unresolved issue is burn. @wouterhar clarified that miners were aware of what was happening during the v2 transition and that there was no major pushback from the mining side. From the holder side, burn looked alarming without enough visible context. That is why communication still matters. @zeussubnet does not look like a fraud or a broken product. It looks like a team that has built more than it has explained. @wouterhar committed to dashboards,investor updates, and more public communication. Those now need to ship. @zeussubnet is building a real B2B product in a real market with a structure for continuous improvement that competitors can’t easily replicate. Communication has to improve. But this AMA changed the picture for me.

English
1
0
13
1.2K