quinn

And if your hardness does not wish to flash and cut through, how can you one day create with me?

@iamgingertrash How do you think that local models can compete with that scale?

Every software startup ultimately decimated You do not understand how dire it looks if you make software, yet

10T class models begin the unemployment wave 10T class models served efficiently get unemployment to ~10-15% 100T class models probably rearchitect human economies It’s surreal to see this happen so fast

@iamgingertrash Do you think this ceasefire affects your call for the bottom or do you think it will hold?

I think that we can call ipfs a success whenever torrenting dies

@DanielleFong Jesus Christ

i need to be monitoring my usage better

@levelsio Photonic matmuls are fucking based. Doesn’t do anything without nonlinear activations. Shkreli’s company doesn’t have any published research or any photonic expertise. There are several other companies who have way more experience in the field

Shkreli has a great track record calling out bullshit so when he thinks something is promising it probably is because he's highly critical

@peashootercat Oof yeah it did

Camming for 40 year old men when you were like 14 does some damage to the brain

@cexscan @BitcoinNewsCom What are you fucking talking about it’s an encrypted mesh network. Block has no elevated permissions Read the code lol

@BitcoinNewsCom Sounds like another way for Jack to collect more data. Curious to see if they'll actually decentralize this properly or just slap "decentralized" on it. We'll watch the bid on that.

Jack Dorsey's Block just launched mesh-llm. It's a decentralized, peer-to-peer inference network for open source AI models. The idea is to pool spare GPU compute across machines to run models too large for any single device. Rather than using a centralized cloud, it's just nodes gossiping over a mesh. Your spare GPU becomes part of a distributed AI network that anyone can use. It uses Nostr for node discovery and the whole thing is MIT licensed and built on llama.cpp. It has the same open and permissionless philosophy as Bitcoin, without a central server that can be switched off.

God I wish I had more compute

@TAOTemplar Bro actually explain to me entirely sincerely why you think the actual token has value. Not ideals, not speculation, why do you think Tao has actual value?

Why is Bittensor so cultish? Here's my answer: I Lost All Hope in Politics… Bittensor (TAO) Restored It.

inspired by @fiveoutofnine's post the other day i made bad apple out of prime numbers

@krzyzanowskim gitlawb.com/node/repos/z6M… Rip your moat

@krzyzanowskim Closed source lol

I reimplemented "claude" CLI with codex and gpt-5.4-high. It cost $1100 in tokens, and is 73% faster and 80% lower resident memory during sustained interactive use. It is very easy to reverse claude from npm distribution, then reimplement is 1:1. It is indistinguishable from the Anthropic version to the every header and analytics it send back github.com/krzyzanowskim/…

@martinez_j7902 @KimDotcom What am I looking at here? I see a delay between the force on the microphone and Charlie getting shot. Am I wrong?

Charlie Kirk's Assassination. The official FBI story is falling apart. The FBI keep trying to push the narrative but the online videos of the shooting are exposing the lies. A modified mic was the weapon that kiIIed him possibly detonated by his bodyguard much like the 'Israeli pager bombs' used in the Lebanon attack last year.

If Kash Patel isn't gone tomorrow Trump is a co-conspirateur in the assassination of Charlie Kirk.

@jpmorais80 What are the jobs you think are going to be in zk? I’m getting into them recently and curious about the prospects

Zero-knowledge proofs are one of the hardest subjects I’ve ever studied, and I used to study black holes and holographic quantum field theory. Part of the difficulty, I think, is that they sit at the intersection of cryptography and computational complexity, which demands a background that few people have - me included. On top of that, many of the ideas are quite abstract. Learning how to construct a SNARK like Groth16, PLONK, or a STARK is not that hard. But truly understanding everything involved takes time, reading, and reasoning. You eventually get there, but it takes patience. A PhD takes at least four years. Expecting to master zero-knowledge in a single year would be too optimistic.

New meta just dropped?

@peashootercat Trikz servers in 2026

STOP PLAYING CS2 START PLAYING CSS STOP PLAYING CS2 START PLAYING CSS STOP PLAYING CS2 START PLAYING CSS STOP PLAYING CS2 START PLAYING CSS STOP PLAYING CS2 START PLAYING CSS

@iamgingertrash @btwphones What’s your timeline on chips?

@btwphones Yes Part of the reason I’m in China is to solve such problems Requires novel ideas around heat dissipation Also, our next gen chips will be ~5x more efficient than nvidia

The wave of model capabilities isn’t slowing down, at all There is a very good chance of 20% unemployment eoy 2027 Stockpiling compute is a good idea Another B200 will be purchased by us

We’re going to get schizophrenic animals because of this no?

@JesusMartinez @markjeffrey It’s kind of retarded to call Tao’s utility as making subnets or mining because what do either of those net - making Tao lol Subnets should better utilize buybacks or payment with staked alpha. Right now it’s entirely ponzinonic

Stillcore Capital wants to own 1% of all TAO. @markjeffrey told me in a recent episode on my channel he thinks $3,000 per coin by December is "conservative." Here's the full breakdown from my 40 min interview with Mark Jeffrey. Bittensor just had its first halving in December 2025. Daily emissions cut from 7,200 to 3,600 TAO. Only 51% of the 21 million supply has been emitted. Mark's comparison: Bitcoin went from $250 to $10,000 in the same post-halving window. But TAO has something Bitcoin never had. Utility demand. • You must spend TAO to create subnets • You must stake TAO to mine • You must stake TAO to get subnet tokens • Only 20% is currently staked in subnets. That number is going to 80%+ The catalyst stack in March alone: • Templar trained Covenant-72B. 72 billion parameters. 70+ contributors. Commodity hardware. Largest decentralized LLM training run in history. • Jensen Huang endorsed decentralized AI training on the All-In Podcast. TAO jumped 17%. • Grayscale filed an S-1 for a TAO ETF. First US ETF for decentralized AI. • Targon launched confidential compute with Intel TDX. Private AI inference at 1/10th the cost. Ridges (Subnet 62) is scoring higher than Claude and Cursor on coding benchmarks. They spent less than $1 million. Anthropic spent billions. Chutes is already the majority inference provider on OpenRouter. People are using Bittensor products without knowing it. 90% cheaper than centralized cloud. Mark put it simply. AOL lost to the open internet. Windows lost to Linux. Gold is losing to Bitcoin. Big AI is next. The math at $3,000 TAO: 334 coins to become a millionaire. About $95K at today's price. $63 billion market cap. Solana territory. Solana does smart contracts. Bittensor does intelligence. Not financial advice. But I've never been more convicted on an AI play in crypto.

@menonfilmpod Where can I watch this tv show I can’t stream it anywhere

Nirvanna the Band the Show (2017) dir. Matt Johnson My Dinner with Andre (1981) dir. Louis Malle