
Building @Halcyon - We want to be the world’s biggest energy information platform ⚡️
San Francisco, CA
Joined May 2009
- Tweets10.8K
- Following3.7K
- Followers63.4K
- Likes33.9K
bruce retweeted
We’ve priced our IPO at $83 per share, with trading expected to begin on Nasdaq tomorrow.
anthropic.com/news/anthropic…
Legal disclaimer: A registration statement relating to the Class A common stock was declared effective by the Securities and Exchange Commission on November 18, 2026. The offering is being made only by means of a prospectus. Copies of the final prospectus may be obtained, when available, from: Goldman Sachs & Co. LLC, Attention: Prospectus Department, 200 West Street, New York, NY 10282, by telephone at (866) 471-2526, or by email at [email protected]; or Morgan Stanley & Co. LLC, Attention: Prospectus Department, 180 Varick Street, 2nd Floor, New York, NY 10014, or by email at [email protected]
bruce retweeted
Can open-weight models write challenging Verus proofs for TLA+ specs?
We tested @NVIDIAAI's Nemotron 3.5 Super VL and the answer is...absolutely, yes!
Our fine-tuning boosts performance from 43% to 73% on the hardest benchmark.
Read our latest report on novel datasets, fine-tuning 3.5 Super VL, and bridging the gap to real-world software verification:
reasonable.io/blog/post-trai…
bruce retweeted
Dense with new stuff that moves the formal verification frontier forwards:
- new agentic fine-tuning datasets, production tested with @NVIDIAAI
- our first demonstration of post-training quantization, making a fast model faster and fitting on a single H100/H200 GPU
- thousands of proofs of real-world software and systems, meaning better eval and benchmarks
and another win for open-weight models.
awesome work @ReasonableIO team!
Can open-weight models write challenging Verus proofs for TLA+ specs?
We tested @NVIDIAAI's Nemotron 3.5 Super VL and the answer is...absolutely, yes!
Our fine-tuning boosts performance from 43% to 73% on the hardest benchmark.
Read our latest report on novel datasets, fine-tuning 3.5 Super VL, and bridging the gap to real-world software verification:
reasonable.io/blog/post-trai…
bruce retweeted
Our team got early access to Nemotron 3.5 Super VL a few weeks ago and decided to sprint a marathon’s distance to land this project: a new dataset of over 3000 challenging Verus proofs of temporal logic (TLA+) properties, battle tested to post-train a pre-release model for agentic software verification.
Building a company is not stress-free, but this amazing team and the intensity of the work have been such a reward.
reasonable.io/blog/post-trai…
Replying to @bensmithrugby
Games won:
All Blacks 1-2 Springboks
The only stat that matters
bruce retweeted
Chat Bietjie Ball 2026 Ep007 youtu.be/MYNrvWVI8kg?si=eHL1…
bruce retweeted
The web is searchable. Podcasts should be too.
Introducing Radar by @particlepro_.
Radar lets you search and spot trends in over 130,000 actively transcribed podcasts, with ~20k new episodes added daily.
Millions of hours, billions of lines: all fully searchable, entity extracted, and rich with metadata.
Set up alerts to your Slack, email, or webhook, to get notified when the entities or people you're tracking are mentioned.
Radar is powered by the Particle Podcast Intelligence API. With Radar, we're now bringing this suite of services to humans, too.
bruce retweeted
Anthropic needs to solve the computer/power problem or they will be the Friendster of the AI era.
I just ran a semi-complicated stock screening prompt on all four major AIs: Grok, Gemini, ChatGPT and Claude.
The first three returned comparable results. Claude refused to do the work.
Not the way to win guys…
Cerebras just had the biggest IPO of the year.
Founder @andrewdfeldman says the 3 most important things he had to convince investors of while doing the roadshow were that demand for inference is going to 1,000,000x, the GPU isn't the only way to do compute, and that the CUDA moat is overstated.
What he said:
"Jensen said some time ago on @altcap's podcast that the demand for inference will grow by a 1,000,000x, and nobody believed him. And at the same time, you saw Sam Altman displaying real vision and going out and trying to lock up huge amounts of compute, memory, data centers, and power, because he saw it too."
"[We tried] to share what that means — what exponential demand means. And that we're still so early, and yet the demand for AI compute is overwhelming."
"The other thing is that there are lots of ways to do this. The GPU isn't the only way. You've got TPUs, Trainium, and us. There are lots of different ways to build a solution here."
"And finally — the notion that CUDA is this grand lock-in is overplayed. Gemini 3, which is an excellent model, was trained on TPUs with no CUDA. The Anthropic models were trained on Trainium with no CUDA. Some of the best models, some of the most interesting things are being done without CUDA. And that lock-in might be overplayed."
$CBRS
Halcyon Curated Alert: FERC Proposed LNG Projects
This alert tracks every proposed LNG project before FERC — from new uprates and liquefaction stages to full expansions — so you catch filings, engineering requests, and air quality reviews the moment they drop.
Do you need to be in the know for Proposed LNG projects or Approved LNG projects?
Subscribe now! Link in comments👇
Halcyon Curated Alert: Utility Mutual Aid Agreements
Track how mutual aid agreements are evolving across state commissions — from shared crews and equipment to more proactive, tech-enabled disaster response.
Get updates on new agreements, participating utilities, and how frameworks are adapting to more frequent outages.
Subscribe now. Link in comments 👇
bruce retweeted
It feels ironic that stacking mac studios is the cheapest way to run frontier AI today.
But it actually makes sense. Nothing else beats the memory unit economics of Apple Silicon.
Memory and memory bandwidth (not FLOPS) is what matters for local inference. Everything runs at batch_size=1. @karpathy talked about this at his AI SUS keynote: personal computing v2 is coming.
Just look at the numbers:
M3 Ultra: $18/GB, $6.70/GB/s
DGX Spark: $36/GB, $17/GB/s
B200 (DGX): $360/GB, $8/GB/s
Memory for B200 is 20x more expensive. Memory bandwidth for DGX Spark is 2.5x more expensive.
If DeepSeek V4 is >1T parameters (as it’s rumoured to be), by far the cheapest way to run it will be Apple Silicon.
M5 Ultra will likely push this further.
NVIDIA has completely missed this segment of the market.
bruce retweeted
My friend @NatBullard wrote a great piece here.
Five hyperscalers are on track to spend $700B+ in 2026, up $300B from last year. That’s before counting partners, utilities, fiber, and supporting infrastructure.
As a share of GDP, it might be the largest single-year capital commitment in U.S. history outside the Louisiana Purchase!
The closest historical analogs include learnings.
Like shale E&P that were short-lived assets and forced efficiency gains over time.
Like Apollo, the hardware depreciates fast but the learnings can compound broadly, and the durable infrastructure (data centers, power contracts) will outlast the compute (GPUs) inside it.
And like the Louisiana Purchase, the full value may only become clear when technologies and use cases that don’t yet exist arrive to fill the space being created.
Replying to @ShanuMathew93
We’ll be at #CERAWeek2026! Our CDO, Alex Klaessig, shares Halcyon’s story and what AI actually means for energy ⚡
Halcyon | AI — Existential Threat or Unfair Advantage?
When? Mar 25 | What time? 9:30am | Where? Pod 1
More details about the session: ceraweek.com/en/program/halc…
Let us know in the comments or drop us a DM if you’ll be around, we’d love to see you!
⚛️ Halcyon Curated Alert: Nuclear Fusion
This alert tracks where nuclear fusion appears in filings — including usage plans, cost discussions, technology pathways, and the manufacturers and companies involved.
If a utility, regulator, or developer is seriously discussing fusion, you’ll know.
Subscribe now with just one click 👇
app.halcyon.io/alerts/subscr…
Halcyon Curated Alert of the Day: Thermal Energy Networks (TEN)
Surfaces key developments related to thermal energy networks, pulling insights from filings, pilot projects, policy signals, and deployment approaches into clear, scannable highlights.
Subscribe now with just one click 👇
app.halcyon.io/alerts/subscr…
Halcyon Curated Alert: Wildfire Mitigation Plans 🔥
Monitors wildfire mitigation plans and related filings that affect grid safety, reliability, and utility operations — particularly relevant for fire-prone regions.
Subscribe now with just one click 👇
app.halcyon.io/alerts/subscr…





















