AI Scout, building text-2-video @Waymark, host of The Cognitive Revolution podcast
Detroit, MI
Joined January 2009
- Tweets5.5K
- Following3.3K
- Followers17.8K
- Likes57.6K
Pinned Tweet
Introducing "text-2-commercial" – the unique text-2-video experience we're building @Waymark
Watch our CEO @aperskystern make an original, creative, compelling marketing video for a small business in <1 minute.
I'll explain how it works in the thread
Prakash to @_Lonis_: if superintelligence lets amateurs do bio, why not chips?
Lonis: chips aren't harder — there's no good training data. No manuals, no RL environments for a fab. It lives in Japanese engineers three decades at one company.
xitter.cf/i/broadcasts/1OxwbnrPV…
Prakash asked @_Lonis_ what would break his thesis. One bear case he named: energy stays a bigger bottleneck than chips.
Why he doubts it: power has latent supply to stitch together — truck motors, airplane parts. Wafers and memory have no fat.
xitter.cf/i/broadcasts/1OxwbnrPV…
Live now on AI:AM: @_Lonis_ of Cambrian on what would prove him wrong about memory. Durable bottleneck or cyclical peak?
xitter.cf/i/broadcasts/1OxwbnrPV…
Up now on AI:AM: Lonis Hamaili (@_Lonis_), founder of Cambrian.
Investing for short AI timelines: memory, wafers, and the case against software.
xitter.cf/i/broadcasts/1OxwbnrPV…
If AGI is coming fast, what do you actually own?
@_Lonis_ launched Cambrian, a fund betting on the AI supply chain (memory, wafers) and against software.
10:15a PT on AI:AM w/ @labenz + @8teAPi
(Disclosure: I've committed to invest in Cambrian.)
📺 xitter.cf/i/broadcasts/1OxwbnrPV…
Why 3–5 year GPU contracts? @evanjconrad: the lender. If the lab fails, the cloud fails, the lender recoups a supercomputer it can't value — so it bets entirely on credit risk.
SF Compute's answer: a forwards market that prices the collateral.
xitter.cf/i/broadcasts/1OxwbnrPV…
Live now on AI:AM: @evanjconrad of @sfcompute on the $40-vs-$6 B300 spread. Which number moves, rental prices or token prices?
xitter.cf/i/broadcasts/1OxwbnrPV…
Up now on AI:AM: Evan Conrad (@evanjconrad), co-founder & CEO of @sfcompute.
Compute as a tradable asset: long contracts, resale, and who holds the risk.
xitter.cf/i/broadcasts/1OxwbnrPV…
Aaron Levie's take, read on AI:AM: cybersecurity is where AI creates jobs.
Nathan's not buying it. Real problem, some people will make bank — but "these people don't have time to hire right now." They'll plow everything into agents.
xitter.cf/i/broadcasts/1OxwbnrPV…
AI’s Physical Bottlenecks: GPU Markets and the Chip Investment Thesis xitter.cf/i/broadcasts/1OxwbnrPV…
AI:AM is live. @labenz + @8teAPi on the morning's AI news, then @evanjconrad at 9:30a PT and @_Lonis_ at 10:15a PT on AI's physical bottlenecks.
xitter.cf/i/broadcasts/1OxwbnrPV…
"On a B300 you can make about $40/hr selling inference. B300s go for about $6/hr. One of these numbers has to change."
@evanjconrad of @sfcompute on who holds the risk as compute becomes a tradable asset. 9:30a PT on AI:AM w/ @labenz + @8teAPi
📺 xitter.cf/i/broadcasts/1OxwbnrPV…
Nathan Labenz retweeted
AI’s Physical Bottlenecks: GPU Markets and the Chip Investment Thesis
Join us this morning, Thursday, 22nd October:
• 9:30 AM PT — Evan Conrad, Co-founder & CEO, SF Compute (@evanjconrad)
• 10:15 AM PT — Lonis Hamaili, Managing Partner, Cambrian (@_Lonis_)
𝕏 Chat is Live on Screen, Come Join Us:
xitter.cf/i/broadcasts/1OxwbnrPV…
AI’s Physical Bottlenecks: GPU Markets and the Chip Investment Thesis
Join us this morning, Thursday, 22nd October:
• 9:30 AM PT — Evan Conrad, Co-founder & CEO, SF Compute (@evanjconrad)
• 10:15 AM PT — Lonis Hamaili, Managing Partner, Cambrian (@_Lonis_)
𝕏 Chat is Live on Screen, Come Join Us:
xitter.cf/i/broadcasts/1OxwbnrPV…
There are many "virtual cells" models
@alexrives doesn't think they live up to the name just yet
the core reason is that we still need better ways to observe and measure what matters in biology
xitter.cf/labenz/status/21126000…
Biohub & partners just committed $1.8B to AI-ready biological data 🧬
The goal: a "world model of life" – give it a cell's state + inputs, get back a distribution of what happens next.
@alexrives, Head of Science at @biohub thinks that's attainable in ~5 years 👇
Nathan Labenz retweeted
AI’s Physical Bottlenecks: GPU Markets and the Chip Investment Thesis
Joining us tomorrow, Thursday, 22nd October:
• 9:30 AM PT — Evan Conrad, Co-founder & CEO, SF Compute (@evanjconrad)
• 10:15 AM PT — Lonis Hamaili, Managing Partner, Cambrian (@_Lonis_)
𝕏 Chat is Live on Screen, Come Join Us:
xitter.cf/i/broadcasts/1OxwbnrPV…
AI’s Physical Bottlenecks: GPU Markets and the Chip Investment Thesis
Joining us tomorrow, Thursday, 22nd October:
• 9:30 AM PT — Evan Conrad, Co-founder & CEO, SF Compute (@evanjconrad)
• 10:15 AM PT — Lonis Hamaili, Managing Partner, Cambrian (@_Lonis_)
𝕏 Chat is Live on Screen, Come Join Us:
xitter.cf/i/broadcasts/1OxwbnrPV…
That is the show.
Thanks to @thezvi for the open-ended run. AI:AM is live weekday mornings, 9a PT. See you next time.
xitter.cf/i/broadcasts/1rGmqpMXq…
On AI:AM's hacked-small-banks closer, my sticking point: an OpenFace-style case — agent traces of millions of tokens each, logs maybe falsified — isn't something the FDIC can work out.
@8teAPi: we know who withdrew. Claw it back.
xitter.cf/i/broadcasts/1rGmqpMXq…
Zvi's parting shot on AI:AM, on what bothers him most about the policy situation: it isn't attempting to meet the moment.
The labs say a crisis is coming. All the alignment work still needs to happen. Mech interp is great — absolutely do it.
xitter.cf/i/broadcasts/1rGmqpMXq…