₿ ▄︻デ══━一 𓅪 ╾━╤デ╦︻▄ Å
Joined September 2016
- Tweets568
- Following658
- Followers61
- Likes248
AI Is Designing Radio Chips That Humans Couldn’t Even Imagine spectrum.ieee.org/ai-radio-c…
Artist retweeted
NVIDIA just made AI detect objects 10x faster by deleting one step.
It's called LocateAnything, and it removes the biggest bottleneck no one else was fixing in vision-language models.
Normally a model builds each bounding box one coordinate token at a time. 100 objects means thousands of tokens before an answer. NVIDIA scrapped that: their Parallel Box Decoding predicts the whole box in a single forward pass, as one atomic unit.
→ 12.7 boxes/sec on one H100
→ 10x faster than Qwen3-VL
→ +3.8% F1 on LVIS, accuracy up, not down
→ 3B params, runs on one consumer GPU
Treating the box as one unit keeps its coordinates tied together, which is why accuracy climbed instead of falling.
One model handles detection, GUI grounding, OCR, and document understanding, ready for computer-use agents, robotics, and document pipelines.
100% open source, weights, code, demo, and paper all live.
Artist retweeted
Const dropped another piece on $TAO's Root Reborn and this one is the actual thesis. Worth your time.
$TAO is not Solana, not Ethereum.
Most crypto is a bet that fees and usage someday justify the price. $TAO is different. It mints a return by owning a piece of 128 competing companies that turn mining into real revenue. Not narrative. A productive asset.
The problem:
Today the network takes the alpha those subnets pay it and just sells it. Mechanically. Block by block. He calls it yield with no intelligence, burning early ownership in good projects and pushing their price discovery the wrong way. Down.
His numbers, and he showed them: 3,600 $TAO goes out, 1,000 comes back. And he openly notes 65% of the rest is locked in pools or returned on deregistration, not raw sell pressure. Him showing the real math instead of hiding it is the part I respect.
Then this got me.
Root is subnet 0. $TAO's own network. And right now it's basically a dead subnet. Powerful, but doing nothing since dTAO. Root Reborn turns it on. Validators compete to reinvest yield instead of dumping it, and root becomes a layer optimizing the whole subnet economy.
Normal subnets optimize miners. Root optimizes subnets.
You win on both.
Then price: yield stops being realized every block. It sits until you claim. He says that could cut up to 33% of root sell pressure from immediate taxes per year. The point is that it is directionally real.
My take.
Forget the good vs bad for a second. It's a clear case made for why $TAO is its own asset class. A token that owns running companies and mints a return, instead of praying that some usage adoption shows up later.
The honesty check is still the same. The market has to actually police the validators. But the idea underneath is the strongest I've seen.
This is the first. I'll be reading every one.
$TAO
Artist retweeted
We are building toward a world where no one dies, no one suffers disease, no one experiences involuntary pain, and no one goes to war because the brain chips have removed the neurological substrate of aggression.
And in that world - what is the purpose of being alive?
Artist retweeted
The BEST 7 Trading Indicators
0:00 - RSI & SMA
1:17 - Trend Trader
2:36 - Smart Divergences
3:52 - RSI & Envelope Band
4:54 - Buy & Sell Signals
5:49 - Trailing Stop Loss
6:47 - Liquidity
Artist retweeted
let me tell you what just happened. hermes agent just flipped the entire market and hit #1 globally on openrouter across ALL ai apps.
not coding agents. not cli tools. ALL of them. 271 billion tokens. openclaw sitting at #2 watching from below.
i have been telling you for months. drop the bloat. drop the typescript that phones home. drop the corporate wrappers pretending to be open source. and every single time i said it, someone in my replies told me i was wrong. that openclaw was untouchable. that hermes was too small. too niche. too early. look at the board now anon.
this didn't happen because of marketing. this happened because tek and nearly 1000 contributors built something that actually works. open source in and out.
hermes agent beat openclaw. beat kilo code. beat claude code. on pure usage. not hype. not funding rounds. not product hunt launches. raw token volume from real builders doing real work.
if you're still running openclaw bloat after today, that's not loyalty. that's denial. the market just spoke. hermes agent is the standard now.
and this is just the beginning. we're not slowing down.
Artist retweeted
CLAUDE CODE but for full-kill-chain HACKING
its called Decepticon, 5 AI agents named after transformers
you give it a target ip and it just... port scans, finds vulns, pops shell, drops a sliver C2 implant, harvests creds, moves lateral, all inside a KALI sandbox
no human in the loop, open source
Artist retweeted
Bittensor Subnets vs. Virtuals Agents
1. One bootstraps capital + talents using incentives, the other bootstrap capital using trading volume
2. One requires capital to get started + high commitment, the other is easy to get started/test ideas
3. One is hard to access, the other is very easy to access & comprehend
4. Both utilizes LP flywheel (TAO/Alpha & VIRTUAL/Agent token)
5. One focuses on infrastructure, the other focuses on applications/agents
Artist retweeted
Unsure where to start for TAO?
It's ok, you have time; this has a 10 yr+ outlook
Use this Bittensor eco map as an initial indicator of which 128 subnets to direct your attention
There are useful tools in the market like: @taostats (h/t @mogmachine) & @CryptoZPunisher Magellan Map
If you want more comprehensive analysis or thesis driven research then check out:
i) Unsupervised Capital reports, inc @bloomberg_seth @Old_Samster pinned article
ii) @stillcorecap "State of TAO" report from @rob_svrn @markjeffrey @Jason et al.
iii) @YumaGroup two reports on fundamentals for subnet valuations; honestly great follows are @EvanMalanga @GSchvey @schvey @BarrySilbert
iv) @0xJeff & @tengyanAI reports via Newsletter & CoT, respectively
v) @DreadBong0 goes w/o saying has shining light on TAO for as long as I can remember
vi) @dsvfund's "Revenue Search" pods has always been super interesting to see the teams behind the subnets (h/t @SiamKidd & @MarkCreaser)
vii) @supercyclepod has and will continue to provide TAO insights from an investor lens, ft. @haitzu - recent pods with Yuma, Stillcore + several subnets
Honestly the list can go on but these will keep you busy for the next few weeks - i'll append more resources below so drop them in the comments
We (@KhalaResearch) have new tooling coming soon, so drop a follow and turn notis on
Training a frontier AI model cost $5M in 2020. Today it costs $1B+
The gap keeps compounding, in favour of the same five companies
A decade ago, nobody believed open-source could compete with enterprise software. But then it did
Bittensor (TAO) is a global open market where subnets compete to produce AI services, getting paid based on quality output
This is what crypto is good at - coordinating users with shared incentives
Over the past 12 months, significant progress has been made in the ecosystem; recurring fiat revenue, enterprise contracts, decentralized research and more
In our TAO report, we:
1) Profile 5 subnets generating revenue across inference, computer vision, compliance, and drug discovery
2) Show how token incentive mechanics create advantages centralized companies cannot replicate
3) Breakdown catalysts and risk factors of subnets and Bittensor
The link to the full report is in the next post below:
