←
Greg Schvey2,277 posts
Greg Schvey avatar

Greg Schvey

@GSchvey

Tech, markets, history, MMA

◷ Joined May 2022▧ 392 Photos and videos
Greg Schvey avatar
Greg Schvey@GSchvey·Nov 13

Had a similar event in my @tesla Model Y with FSD on this week. It may have saved me from serious harm. In the first video, you can see FSD detect a deer on the highway in the right lane. It seemingly detected it before the car that was in the same lane as the deer and eventually hit it. You can see in the video where self driving turns off (labeled on bottom) because I instinctively hit the brakes as the deer went directly into my lane. In the second video you can see my rear camera view, where an 18-wheeler behind me in my lane was coming in hot. Thank goodness I thought to check and acted quickly, pulling forward and to the right. The third video with the side camera view shows how close the truck was to hitting me. You can see in the attached stills, the truck just barely gets far enough left to miss me about 50+ ft past where I was initially braking. If I hadn't moved both forward and right I would have gotten plowed. Thing is, I believe FSD would have seen that and reacted even better, like in the quoted post. It has 360 context that no human can match. My recommendation is to default to leaving FSD on if you're in a similar situation. Also, huge gratitude to the @amazon driver on 287 near Nanuet, NY on Wednesday who moved into the left lane as quickly as he could to spare me from catastrophe.

Post attachment
0:19
0:16
0:16

My brother’s friend drives a Tesla. This is 3 videos of the same accident. The Tesla, on self drive, sees the approaching car coming from behind at excessive speed and moves out of the way. Driver of speeding car was killed. Amazing technology @elonmusk.

0:26
0:15
0:22
Greg Schvey avatar
Greg Schvey@GSchvey·Nov 12

Always great to see Bittensor creating value beyond the digital world

CADmind (SN42) has joined the Yuma Accelerator. From device cases to drones and robotics, CADmind brings prompt-based generative AI to physical design and production, with 3D models built on Bittensor. $TAO

0:19
Greg Schvey avatar
Greg Schvey@GSchvey·Nov 11

Sincere thanks today - and every day - to our veterans for the sacrifices they made so we can live in the greatest civilization mankind has ever known.

Yuma avatar
⟳ Greg Schvey retweeted
Yuma@YumaGroup·Nov 3

Fugal is joining the Yuma Accelerator. Not every query needs the most powerful model. Sending every token through frontier inference means overpaying for intelligence that smaller models handle perfectly. @fugal_ai (SN70) is building a LLM routing subnet that matches each query with the lowest-cost model capable of answering it correctly. Miners compete to improve the routing layer, continuously optimizing the tradeoff between complexity and cost. Many of the world’s largest companies build these optimizations in-house. Fugal is building them for everyone, turning efficient model routing into an open market on Bittensor. $TAO

0:17

fugal is live on bittensor: subnet 70. the best model for your question often isn't the most expensive one. and which one is best keeps changing: new flagships, new prices, and nobody is paid to keep re-measuring. fugal is an llm router. it reads your question, predicts which model will get it right, weighs that against price, and calls that one model. built to give you frontier answers at a fraction of the cost. subnet 70 is how it stays right. miners compete to train better routers, every benchmark runs inside intel tdx so results can't be faked, and they only earn when their router beats every fixed choice of models at the same cost. so the only way to win is a router that gets more right for less. starting today: • miners can register: github.com/fugal-ai/fugal-su… • scoring starts today • every question, grade and score gets published each epoch try the router: fugalai.com

Gyles | TAO.com avatar
⟳ Greg Schvey retweeted
Gyles | TAO.com@Gylez·Oct 30

$1M+ in week-on-week volume through TAO.com Intents on ALPHA ↔ USDC since launch, and so far, that's come entirely through the simple swap frontend. Incredible early signs of what's to come as we start onboarding more partners through the SDK over the coming weeks as well as enabling iOS to be multi-chain. Successful subnets should ultimately sell themselves through token buybacks and burns that offset miner sell pressure, creating self-sustaining economies and powerful flywheels. Our job is to build the distribution rails to make these accessible to every market.

Yuma avatar
⟳ Greg Schvey retweeted
Yuma@YumaGroup·Oct 21

The SEC’s proposed Regulation Crypto Assets is meaningful progress toward a clearer framework for digital assets, and we support it. Work remains on how token issuers can transact in secondary markets. Yuma submitted a comment letter today proposing a framework based on disclosure and discretion, with objective standards designed to give issuers and market participants greater clarity.

Post attachment
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 14

This is long-term capital going into subnets at scale. These funds have a min 12mo holding period and most of Yuma's investors have an investment horizon far beyond that. That's capital directly supporting Bittensor's builders, and $TAO in general, for the long haul.

At the close of September, Yuma Asset Management's funds held 2.0% of Bittensor's total subnet market cap. The same week, AUM reached a new all-time high of $34.4M across our three funds and SMA. $TAO

Post attachment
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 13

Once again we see Bittesnor's Sum of Subnets dominated by small, outlier subnets. SoS is up 5.5% in the last 10 days (3.8x relative to YCX), with 79% of that move from just two subnets - both of which are in the bottom quartile by market cap.

Post attachmentPost attachment
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 8

George Washington was known for making examples of traitorous soldiers via public execution. Most famously, in 1776, Thomas Hickey (in Washington's own guard), was caught trying to enlist Continental soldiers for the British. Washington not only ordered him to be hung, but also commanded all soldiers from four brigades to march him to the gallows and witness the event. The greatest leader our country has ever known employed intentional theatrics to send an unavoidable message: traitorous behavior by American soldiers will be punished with death.

Army mass shooter Nadal Hassan's execution will be public, carried out by 'soldiers with rifles' at Fort Hood: Hegseth trib.al/D0YIKbA

Post attachment
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 7

Notably, basket claims are still a relatively small portion of overall subnet flow. The mid-Sept drop in subnet prices likely had a greater overall impact on validator basket value than claims had on subnet prices.

Post attachment
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 7

The cumulative size of all Bittensor validator baskets has levelled off at ~25K $TAO value since mid-Sept When RR launched, subnet tokens stayed in baskets, reducing sell pressure relative to auto-swaps to $TAO that used to be common. As more stakers began claiming, latent selling pushed value out of subnets into root, which seems to have generally levelled off at 200-500 TAO/day

Post attachment
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 7

Yuma Large Cap Fund up 64.6% since launch, vs 48% for $TAO This fund has done particularly well in the last few months with the increasing concentration of emissions to the highest price (generally largest) subnets

The Yuma Large Cap Fund gained 29.0% in September on a USD basis, as its larger subnet holdings outperformed the broader subnet market. It was Yuma's strongest fund in Q3, up 64.6% vs. 48.0% for bittensor:native. The fund's composition is analogous to how the Mag 7 represent a concentrated allocation to dominant technology companies in public markets. Learn more about Yuma Asset Management funds at go.yumaai.com/utdUl Disclosure in thread.

Post attachment
Gordon Frayne avatar
⟳ Greg Schvey retweeted
Gordon Frayne@gordonfrayne·Oct 6

Building a successful subnet takes more than a good AI idea. I sat down with @LindsMikeStone from @YumaGroup at @ExploitSummit to hear more about how their accelerator helps Bittensor teams launch, grow and build businesses around their technology.

12:49
Greg Schvey avatar
Greg Schvey@GSchvey·Oct 6

The Yuma Composite Fund is up 30.2% since launch vs $TAO down 3.9%. The difference is driven by highly diversified subnet exposure, which reliably generated ~35% annualized yield from emissions over the last year while also directly supporting Bittensor's builders.

The Yuma Subnet Composite Fund rose 26.9% in September on a USD basis, its second-best month of 2026, while $TAO gained 30.9%. Subnet tokens trailed TAO’s recent rally, but over Q3 the fund returned 52.5% versus 48.0% for TAO. Since launch it is up 30.2%, while TAO is down 3.9%. (All changes on a USD basis) The fund provides indexed, weighted exposure across Bittensor subnets, analogous to how the S&P 500 provides diversified exposure to the U.S. equity market. Learn more about Yuma Asset Management at go.yumaai.com/utdUl Disclosure in thread

Post attachment
Trishool | SN23 avatar
⟳ Greg Schvey retweeted
Trishool | SN23@trishoolai·Oct 5

A month ago we showed you the first version of our output guard. 73.9% F1. We said it was the floor, not the ceiling. Two weeks later it was at 78.53%. Today it's at 81.87%, less than a point from SOTA. Every one of those points came the same way. Miners across the network attacked the guard week after week, every successful attack became something it learned from, and each new version came back harder to break. We didn't hide the starting number and wait for a polished result to announce. We showed it to you at the bottom and let you watch it climb in the open, because the climb is the proof the whole system works. The input guard already sits at the top of its class. The output guard is now right at the door. Keep watching. The next update is the one we've been building toward.

Post attachment