Nick Carpinito

@jacobmtucker @cursor_ai Codex + Claude Code extensions for the win

@jacobmtucker @cursor_ai Couldn’t agree more

For the first time in years, I switched back to default VS Code @cursor_ai keeps moving further from what I want: a code editor with search I don't want an agent window by default or another ChatGPT UI. I want to manage code, browse structure, and see changes Like what is this

JTX launching today. The biggest catalyst for Jito, turning them from a Solana backend infrastructure into a distribution controlling frontend. Likely one of the best teams and hence also the most promising attempt for the Solana ecosystem to compete for perp DEX market share at the highest level. I will watch their launch and its adoption over the next few weeks closely, but as long as it's not a complete failure, I would expect their token to reprice aggressively. Probability weighted price target for $JTO: $2.75

@doublezero Because big month?

tell us why you deserve this

1000

Is it unrealistic to have the people who regulate technology understand that technology?

can this be the next trillion dollar narrative? we all need to see how AI compute could change a lot. these are not the AI models; this is the infra that powers them. wall street has started treating GPU compute like a commodity. in May alone, @CMEGroup, @ICE_Markets and AIX moved to launch compute futures. multiple issuers filed ETFs tied to compute price benchmarks. by that, you should know so much is already going on. this is how we need to come in: today's compute markets are still built around cash-settled contracts. you can hedge GPU prices, but you can't verify the compute or reliably access supply. that's the gap crypto can fill. instead of competing with tradFi, we can become the infra layer beneath it. that means building: > verifiable compute, where cryptographic proofs like ZK proofs confirm AI workloads were executed correctly. > decentralized compute networks that aggregate GPU capacity from providers around the world. > onchain settlement and GPU-backed credit, making compute programmable, transparent, and financeable. some projects are already building toward this. @akashnet and @ionet provide decentralized GPU capacity for AI workloads. @Lighter_xyz has listed H100 perpetual markets, while @HyperliquidX supports compute markets through its HIP-3 builder framework. on the verification side, @nockchain is working on verifiable AI inference using cryptographic proofs. individually, these look like separate products. together, they're pieces of an entirely new financial stack. this is a new narrative where crypto becomes the infrastructure behind compute markets. tradFi is already building the future. as a space, we could build the settlement, verification, delivery, and liquidity layers that make the entire market work. this is the financialization of compute, and it's a sector worth paying close attention to.

Another challenger has appeared

Our GPUs are not the same

@ryanconnor Colossus 1 was a disaster. 2 and 3 are more promising tho

Re Grok, after seeing today’s evals, it’s not crazy to think xAI achieves a decisive lead in the frontier model space in the next 12-24 months from a revenue perspective. As Elon has a unique advantage across all three scaling laws. Pretraining & Colossus. xAI’s ability to spin up bleeding edge data centers its unmatched. Colossus 1 was the first coherent 100k GPU cluster. Colossus 2 first GW scale dc. Jensen said that no one else comes close. The confers major advantage in the compute half of the pre-training scaling law, (arguably the most important of the three scaling laws). Plus they probably run into capacity issues far less frequently as users ramp. Now apply Elon’s “algorithm”, and xAI dcs likely run x% more profitably than the competition. Pretaining, RL & Proprietary coding data. Cursor is rumored to have just as much coding data as Anthropic, maybe more. Unique access to data is arguably the largest technical moat in this business. This gives xAI the pretraining and RL data they need to compete. As a side, if coding is in fact the killer use case that enables RSI, and then ASI & AGI, xAI then gets there first. Test-time & the budget to think for longer. Bc Grok is ~90% cheaper than Fable, you can think for longer with a more inteligent model with the same inference budget. So you again win in test-time too.

@Nomaticcap But no insurance

Damn, 7% is getting quite interesting.

@defi_kay_ Goated onboarding buddy. You will be missed

Today was my last day at Blockworks I hope that you've enjoyed hearing from me on 0xResearch and Lightspeed during my time as host TBD on what's next for me, but just want to say thanks to all the BW folks for a great few years, and all the listeners for tuning in 🫡

Open-source AI is the new cypherpunk agenda

solana:Grass7B4RdKfBCjTKgSqnXkqjwiGvQyFbuSCUJr3XXjs just wrapped token holder call #2: Back in November they reported revenue with no cost structure. Today they delivered. Revenue: - FY25 landed at $17M - Their December materials guided Q4 2025 to $12.8M, which implied ~$20M for the year. H2 2025 came in at $14.3M, so Q4 landed near $10M - The FY26 projection is $65-75M. - H1 2026 is already booked at ~$17M - That leaves H2 2026 to carry ~$53M Profitability: - ~$2-3M/mo opex annualizes to $24-36M against a ~$34M revenue run rate - The 2025 infra investment cut >$1M/mo in opex - Team indicates they are profitable Team also confirmed no equity ever raised, all funding through the token, with Wynd Labs holding the code and revenue flowing to the network. Disclosure is catching up, but the 2026 story now hinges on the H2 ramp.