Bittensor’s first crowdfunding platform. Discover new teams. Pledge TAO. And get liftoff 🚀 Open for all: TG @ https://xitter.cf/t.co/16ayiZ35IL From @macrozack 🫶
Joined September 2025
- Tweets369
- Following144
- Followers3K
- Likes1.8K
bitstarter retweeted
Apex was built to hillclimb intelligence.
Our agent harness allows anyone to launch a live incentive mechanism in subnet 1 without writing a line of code. We've reduced the time to create competitions from weeks to minutes. Once live, Apex competitions are mined by a swarm of agents from around the world.
We've solved access to decentralized intelligence.
Now it's time to ask the big questions.
We're very excited to partner with @bitstarterAI as the home of many teams with big ideas.
Join us on Friday for a live community call and AMA.
Apex Accelerator: SN1 x Bitstarter
Great concept. But will it work on mainnet?
Now, our teams can find out - testing their incentive mechanism and getting results from miners on @Apex_SN1 - all before they launch.
Full reveal this Friday 16th with @macrocrux - coming soon⚡️
bitstarter retweeted
Cascade’s model outperforms Toto 2.0 4M on TIME MASE and TIME CRPS:
- With just 15.4% as many training tokens
- Using only synthetic miner-generated data
- In only 58 days
Let’s break down what TIME MASE and CRPS are, and why they matter.
bitstarter retweeted
One week ago, I introduced @carbonphysicsai on stage @ExploitSummit - who we placed in SN116, now ranking as a top 5 subnet by price on Bittensor.
One year ago, I launched @bitstarterAI - promising building a community-backed incubator capable of powering teams to the top.
Nine launches later, we're finally getting there.
Thanks to everyone who pitched, pledged, and judged (whether on-stage or otherwise) - none of this would have happened without you.
🤖 Made with AI
bitstarter retweeted
Breakdown of @markjeffrey hashrate on @cascade_sn91
Aim: To tackle time series modelling.
LLMs have easily grown due to the abundance of data.
Time series data is limited due to having one path through history.
Cascade aim to build both the time series model and produce the synthetic data that it can be trained on.
- Examples of the effects on time series data include
- Trends / Seasonality / Main Reversion / Abrupt news.
- Cascade can become a universal prediction engine, due to a lot of similarities within the patterns that build up the model the same way different chords can make different songs (a nice comparison might I say by @macrozack )
- The competition is in favour of smaller companies since the playing field is the same due to the amounts of existing data.
- This model can mimic similar prediction subnets like 18, 50 but more as a general prediction.
- powered by @engyai
bittensor:native #bittensor
Hash Rate - Ep. 186: Cascade Prediction SN91
🧙 Guest: @tensorlink_dev @macrozack of @cascade_sn91
00:00 Introduction to Cascade Subnet 91
09:24 The grammar of 'universal prediction'
12:08 Data efficiency and advantages of small models
14:15 Proving models in the field and future applications
17:42 Deployment, API, and community engagement
19:45 Future plans, competitions, and use cases
22:54 Forecasting tools for the Bittensor ecosystem
28:16 Hosting models on SayGM
29:54 Chris Haymes' background
39:51 Integration with Hermes Agent
50:23 Impact of recent emission changes
56:46 Closing remarks and next steps in AI prediction
bitstarter retweeted
I sat down with @willobrien at @ExploitSummit to hear more about @provenonce_ai and their plans to close the assurance gap in a world of agentic AI!
bitstarter retweeted
What gets a Bittensor team into an incubator, and what happens after the yes?
@LindsMikeStone, Kelly Woodward of @CrucibleLabs and @capradavis join @macrozack of @bitstarterAI on the RAO stage to discuss how teams get selected and supported.
bitstarter retweeted
“Access, alpha, and assurance.”
@macrozack explains what @bitstarterAI aims to give new subnet founders: access to experienced builders, insights and hands-on contributions, and confidence that their idea can work. From Pitchtensor at Exploit Summit.
bitstarter retweeted
“You'll see them in real time getting pledges online and also in the audience.”
You decide who to back. @macrozack explains how Pitchtensor teams compete through @bitstarterAI, with a subnet slot and pledged TAO to get the winner building.
bitstarter retweeted
Bittensor can be a force multiplier, but teams have to have the right vision, execution, and support behind them to succeed.
At Exploit, this panel gets inside how incubators decide which teams to fund, what kills promising projects early + what founders need to get right to earn conviction.
With @LindsMikeStone of @YumaGroup, Kelly Woodward of @CrucibleLabs, and @capradavis of @UnsupervisedCap + moderated by @bitstarterAI's @macrozack: luma.com/exploitsummit26
bitstarter retweeted
Six models. Three forecasting modes. One API.
Ephemeris is live: time series foundation models from Amazon, Google, Datadog - and soon from #SN91 itself.
Submit a series ➡️ Receive a forecast. Try it now:
ephemeris.cascade.industries…
🤖 Made with AI
bitstarter retweeted
Name one other pitch competition where:
you invest in the teams
while they pitch
live on stage
...and see their fundraise fulfil in real time.
Only on #Bittensor. Only at @ExploitSummit.
Two machine learning teams. One #Bittensor subnet slot. You decide who gets it.
At pitchtensor, hosted by @bitstarterAI, two aspiring teams pitch to @macrocrux, @KeithSingery, @maxsebti, @EvanMalanga + @willobrien - then you choose the winner in a live crowdfund, online + in the room.
Live from the Exploit stage: luma.com/exploitsummit26
bitstarter retweeted
300M scientific papers. Almost none are structured for machines to reason over.
@PKoellinger + @nashgambit of @DeSciClaims show how miners can extract claims + evidence, validators verify them, and incentives turn the literature into trusted, AI-ready data.
Catch them at Exploit: luma.com/exploitsummit26
bitstarter retweeted
Introducing multivariate scoring on Cascade.
Real-world problems are rarely one line. They move together. #SN91 can now evaluate related time series collectively, producing scores that better reflect the structure of real forecasting problems.
Multiple forecast channels are averaged into a single mean, resulting in one GIFT-Eval-weighted contribution. This preserves the value of multivariate data without allowing problems with more channels to carry disproportionate weight.
Multivariate scoring is now live on mainnet.
bitstarter retweeted
🚨 Subnet Summer AMA x SN111 @DeSciClaims
SN111 Claims is building the canonical claim-evidence record for science on Bittensor. It extracts scientific contributions from papers, links each claim to grounded evidence, and makes research machine-readable for AI. The subnet transforms how scientific knowledge is incentivized, validated, and accessed.
🕓 Thursday, September 17th at 4PM BST, we're hosting a live AMA with the Claims team.
We'll cover:
- What SN111 Claims is and why making science machine-readable matters
- The DeSci opportunity: $3 trillion of knowledge locked in human-readable papers
- How miners extract claims, how validators audit for integrity
- The product roadmap: turning a knowledge graph into a revenue-generating API
If you're interested in decentralised science, AI-driven research infrastructure, or how Bittensor tackles knowledge accessibility, this one's for you.
📍: t.me/subnetsummerTG
bitstarter retweeted
Cascade just changed how miners compete.
🔥 No more heat screen. Every challenger trains at full budget, so nobody gets cut before they’ve had a real shot.
🔭 Scoring now spans three horizons: 64, 256 and 720 blocks, pushing miners to optimise across short, medium and long-range forecasts.
👑 A 1% edge can now dethrone the king, intensifying the competition.
1️⃣ One entry per generator, rewarding genuine model diversity, not the same model with different settings.
#SN91
bitstarter retweeted
Ep. 99 - Philipp Koellinger & Christian Roessler
Philipp and Christian are building Claims @DeSciClaims Subnet 111
Timestamps
0:00 - Highlights
1:48 - Proof of Pitch to Mainnet in Three Months
2:50 - Why Economists Find Bittensor Fascinating
3:42 - Citations as Science's Reward Function
4:29 - The Replication Nobody Would Publish
6:53 - How Much Published Science Is Wrong
8:26 - The AI Echo Chamber vs. Human Work
9:54 - Making 300 Million Papers Machine Readable
15:11 - Reproducibility vs. Citations
15:46 - Turning Papers Into a Knowledge Graph
16:20 - Why LLMs Hallucinate Citations
21:21 - Longer Context Windows Make This Obsolete?
22:56 - Coverage, Extraction Quality & Evidence Quality
25:16 - What Counts as a Substantive Claim
26:42 - What This Is Worth to Working Academics
27:37 - Academia Admits the System Is Broken
29:26 - How Do You Actually Find the Truth?
32:06 - Truth as Probability, Not Zero or One
33:33 - Logical Proof vs. Physical Replication
35:44 - Correlation vs. Causality
36:28 - Ground Truth From Replication Studies
38:12 - Codifying Claims to Surface Contradictions
40:14 - Can You Trust an LLM as a Judge?
42:48 - Synthetic Checks: Truth Is Sparse
45:40 - How Much Miner Variance Is Desirable
46:55 - Why Bittensor Instead of a Closed Company
49:50 - Success, Revenue & the Palantir Comparison
51:47 - The Flood of AI-Written Papers
54:36 - Advice for Discerning Truth
bitstarter retweeted
Cascade is building the full stack for time-series intelligence. Not just pretraining models.
1️⃣ SN91 runs the research competition.
2️⃣ TSBench-Forge provides the evidence
3️⃣ Ephemeris provides model access and routing
4️⃣ Gnomon brings that capability into agent workflows
Over the next few weeks, we’ll begin opening access to our benchmark, the API for time series models (including Cascade's model), and the agentic harness.
bitstarter retweeted
Who will become the next great AI team on #Bittensor?
Pitchtensor is @macrozack's answer: two machine learning start-ups, competing live on stage at Exploit.
And the winner will be decided by you - via @bitstarterai's crowdfund: luma.com/exploitsummit26
bitstarter retweeted
On GIFT-Eval: Cascade 4M ≈ Salesforce 91M
One of Cascade’s 4M-parameter models has effectively matched Salesforce’s 91M-parameter Moirai-Base across the full GIFT-Eval benchmark, while slightly outperforming it on CRPS (lower is better).
Cascade: 0.8993 MASE · 0.6009 CRPS
Moirai-Base: 0.9010 MASE · 0.6095 CRPS
Moirai-Base trained for 1M steps on a curated real-world corpus. Ours started from random initialisation and trained for just 150k steps using synthetic data alone.
We deliberately held the Toto2 backbone at 4M parameters, making the data generator the main variable between Cascade runs.
This is our first strong signal that the data really is the lever.
For context, Moirai-Base is an earlier-generation TSFM from Salesforce, not today’s frontier.
Inspect the public checkpoint here:
huggingface.co/tensorlink-de…