The Scuttlebutt
Subscribe
Sign in
Home
Notes
Chat
📰 Editions
🧭 Topics
🎙️ Sources
🗣️ People
📈 Investments
💡 Themes
🗂️ Asset Classes
Archive
About
Bloomberg Tech
Latest
Top
Discussions
Crusoe CEO on water: "roughly the same water footprint as about 10 single family homes"
18 takeaways: Why data center contracts run 15 to 20 years while GPU clusters run months, what SoftBank borrowed to fund OpenAI, and the tax base one…
17 hrs ago
•
The Scuttlebutt
AI extinction fear is "much more science fiction than science" — Google Brain's founder
12 takeaways: OpenAI discloses six cases of its models misbehaving, Databricks' CEO puts the existential risk near zero and the cyber risk very high…
Sep 18
•
The Scuttlebutt
Aidan Gomez: the evaluators being proposed share funders with Anthropic
5 calls: Why an antitrust exemption is the part he objects to, what he wants instead of a lab-written rulebook, and the argument he is making against…
Sep 15
•
The Scuttlebutt
Dave Baszucki: Roblox games will ship as standalone apps on phones and consoles
4 takeaways: What creators earned on the platform last year, the gaming version of a coding harness, and why Roblox is letting its own games leave the…
Sep 15
•
The Scuttlebutt
Parmy Olson: pacing is not slowing, and she does not expect a slowdown at all
4 predictions: What the word pacing is doing in Anthropic's essay, the agreement she expects within weeks, and why a US-China pact on AI is not coming
Sep 15
•
The Scuttlebutt
Peter Elstrom: SoftBank fell about 11% because OpenAI's listing just moved
4 insights: What Beijing's foreign ministry said about the slowdown, why China's labs are not the ones being asked to pace, and how the news reached…
Sep 15
•
The Scuttlebutt
Rayan Krishnan: his index puts models past human AI researchers by August 2027
4 arguments: What the recursive self-improvement index actually measures, the one conflict that would break an embedded evaluator, and why he is more…
Sep 15
•
The Scuttlebutt
Uday Cheruvu: slowing the frontier stretches the demand curve, it does not shrink it
5 calls: Why a pacing agreement is a timing problem rather than a demand problem, what regulation does for the labs that already lead, and why he says…
Sep 15
•
The Scuttlebutt
Metz and Shepard: nobody has committed to ending a training run
4 takeaways: How a resignation turned into an industry-wide call for pacing, what the rival labs actually signed up to, and why Congress is unlikely to…
Sep 15
•
The Scuttlebutt
Mitesh Agrawal: Positron rides Nvidia's rails instead of fighting them
3 takeaways: why the $875M round exists, Oracle's 50 racks already running, and the memory bet behind the next chip
Sep 11
•
The Scuttlebutt
Sid Sheth: no rival will match D-Matrix's memory chip for two years
3 takeaways: why riding Nvidia's install base beats building a rival one, the customer types already lined up, and the memory trick behind it
Sep 11
•
The Scuttlebutt
Mark Gurman: every foldable before the iPhone Duo looks like a beta test
3 takeaways: why the $1,999 price still buys Apple a margin, the $800 Australian premium on top of it, and Apple's first real AI strategy since the Mac
Sep 11
•
The Scuttlebutt
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts