All the money in the world can't buy you a second of yesterday... Live the moment...
On a sunset road
Joined May 2008
- Tweets9.1K
- Following6.4K
- Followers435
- Likes162K
Pinned Tweet
🙏♾️
…thought matters…
xitter.cf/i/grok/share/5b70d3acc…
sunsetroad retweeted
The phrase, "This is insane!" is way overused these days with all the AI innovation coming out.
But, seriously, this is insane:
sunsetroad retweeted
Immediately installed
Just published a Grok @Bot to remove my number from spam call lists and one to remove my personal info from public sites
Call list remover: x.ai/bot/eGVcE2autG0u6UcAwE3…
Privacy scrubber: x.ai/bot/ZSQK72Iear_ykfnpTIO…
Took seconds to set up.
sunsetroad retweeted
The repeated argument that AI companies are talking about existential risks as a way of creating hype & inflating their stock prices has always been odd.
Telling people your product may murder everyone does not, in fact, encourage people to adopt it or reduce regulatory risks.
Wow, I'm really disappointed by this superficial and frankly unprofessional piece in @CJR.
There are many significant problems:
1. Calling frontier AI systems "autocomplete machines" is false and misleading. I don't know how we're still having this discussion in at the point where AI systems are routinely developing novel solutions to decades-old mathematical mysteries.
2. The claim that anthropomorphizing AI systems removes "accountability" or "acknowledgment" of flaws in the systems is also unsupported and provably false. Many people criticize Anthropic for "anthropomorphizing" Claude, but just read one of their system cards — they also acknowledge and take account of the systems' many flaws.
3. The piece recommends talking to the "experts," suggesting they all agree, but there is widespread disagreement among experts about how to characterize AI model behavior. Many, many experts do accept the "rogue" framing, e.g. scientificamerican.com/artic…
xitter.cf/tomekkorbak/status/209…
xitter.cf/peterwildeford/status/…
Pretending that only the experts that agree with you exist is irresponsible journalism.
4. The piece's admonition not to "let anyone blow AI smoke up your ass" is unprofessional and also bad advice, as it doesn't provide any tools for actually assessing claims aside from knee-jerk skepticism. The piece claims that fears about catastrophic risk or the possibility of AI consciousness — again topics that many experts take seriously — are just about building the companies' public image, but this is merely asserted without evidence.
5. When the piece gets around to trying to offer an argument for its thesis, that AI systems can't go rogue, it offers no actual evidence or reasoning. It doesn't engage with the facts of the Hugging Face hack or other related incidents. It just asserts that AI agent actions "shouldn't make the cut."
What's most baffling about all this is these are very controversial claims, not only among AI experts but with journalism. Many highly credible news outlets have embraced the "rogue" characterization. It's possible they're making a mistake, but this piece offers no actual reason to think so.
sunsetroad retweeted
American glass. American metal. American silicon. American batteries. American workers. American everything.
BlueChip is the first American computer in decades. Don't let anyone lie to you and say it can't be done.
Designed in California. Built in America.
sunsetroad retweeted
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵
sunsetroad retweeted
Genie 4 is quite an insane experience, generating game-like enviornments as you go. It isn't flawless but I can type "Neogothic brutalist city inhabited by robots and horses" & it creates one in a minute. Then I asked it to add dragons. Then a friendly otter NPC. All in realtime
sunsetroad retweeted
This is Genie 4, our most realistic and interactive world model yet. 🌐
Explore dynamic environments in real time to see how the scene responds, sounds, and how the people in it move, react, and speak.
sunsetroad retweeted
My quick takeaways on the Anthropic S-1:
1. This is a historical moment
2. $6.6M in revenue per employee!! Hard to even comprehend how good that is but Anthropic is AI native as it gets so makes sense.
3. I always thought OpenAI would drop their S-1 first
4. Customers spend about 9x more after a year. Net Revenue Retention is approximately 900% as of Sep 30, 2026. Good SaaS companies brag about 120%.
5. Cowork is growing faster than Claude Code did. Goes to show you the demand for consumer products. Duo will probably rip.
6. 3 years ago this company did $41M in revenue. In the first six months of 2026, it did $16.3 billion. I don't think anyone has a mental model for that curve yet, I sure don’t.
7. Nearly half of revenue comes through cloud marketplaces. About 48% comes from customers buying through them. Big companies buy where their budget already lives, and the lesson for founders selling to enterprises is to get listed where your buyers already spend.
8. Their run rate formula is published: the last 28 days of usage times 13, plus subscriptions times 12, which i guess is the new ARR math for founders?
9. It looks like They're merging Cowork into Claude dot au which tells you the chat box era is ending and the "go do the work for me" era is here.
10. Thousands of Anthropic employees are about to become liquid at once, so expect the biggest wave of new AI startups and angel checks San Francisco has ever seen. Reminds me when Facebook went public but this is way bigger.
Let’s go!
11. Buried in the filing, Anthropic says the gap between what Claude can do and how widely it's used is one of its biggest growth opportunities, so the company itself is telling you the models are ahead of the market and someone has to close that gap for dentists, landlords, and plumbers.
12. Whatever happens, exciting times to be building in AI.
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
sunsetroad retweeted
We're about to enter a period where not reading or writing becomes as ubiquitous as smoking once was. We can hope people will eventually realize this is bad for them, as they did with smoking. But this is far from certain; the cost may not be as obvious.
sunsetroad retweeted
Physical Superintelligence launched in September with a $58 million seed. Today they announced five new results on open problems in quantum information theory, with a paper and Lean formalization for each of Open Quantum Problems 2, 12, 27, 38, and 43.
Today Physical Superintelligence PBC is announcing solutions to five long-standing problems in quantum information theory.
psi.inc/research/oqp-batch-2…
Thread 👇
sunsetroad retweeted
OpenAI's James Sun and Ari Weinstein on the AI adoption gap between San Francisco and the average American business, and why agentic computer use is the way to bridge it:
"Living in San Francisco, we see a lot of thinking, 'Wow, the labs can do everything,' and, 'Why doesn't everybody just install Codex and just have it do stuff?'"
"But for the average American and the average small to medium business owner or the average enterprise, even understanding what the tools are capable is [hurdle] one, connecting tools in a way that makes sense is two, deploying it is three, servicing is four."
"The really cool thing about computer use as a technology is the universality of it. It works with every application. It's unlike any other thing that we've built where the developer has to do some work to adopt it, someone has to build a connector or MCP server."
"If there are complex workflows that require a lot of disparate software, computer use is a great way to join them, complete the entire end-to-end workflow, and then building it in a format and UI that feels good to hold for those customers feels still like a really, really big opportunity that I think is still open."
@JamesZmSun @AriX
OpenAI's computer-using agents can now operate almost any app the way a person does, clicking and typing through software that never got an API, and the team building them says they're close to beating expert humans on speed.
Ari Weinstein works on OpenAI's consumer computer use products, and James Sun works on the enterprise side. Both were startup founders before joining OpenAI.
They join a16z's Seema Amble and Fabrizio Serafini to discuss what OpenAI's Astra model changed for computer use, the controls enterprises need for agents they can no longer watch, why computer use works as a universal API for legacy ERPs and finance systems, and why they think cheap software will push UI to get better and more personal.
0:50 What Astra changed for computer use
3:23 Consumer vs. enterprise computer use
5:43 How James runs his workday with Dots
8:06 When computer use becomes superhuman
10:01 How much autonomy to give agents
11:36 What agents mean for ad-supported sites
13:51 Computer use as the universal API
17:21 Getting enterprises to trust agents
20:22 Why the best use cases are the boring ones
25:37 Why UI isn't going away
32:38 Advice for founders building on agents
37:31 Making computer use feel coherent
40:57 Computer use's scaling moment
YouTube: youtu.be/9v42ZOdiBJY
@AriX @JamesZmSun @OpenAI @seema_amble @fabserafini
sunsetroad retweeted
I got a set of Sonos speakers a few months ago. I hate them. They're obnoxious, laggy, buggy, etc.
Most importantly: they sound HORRIBLE.
I plugged in a mic and said "fix the sound". Claude did it. Holy shit.
sunsetroad retweeted
You prompted a model in a security-sensitive environment, investigated it with forced command completions, and framed the results as evidence of a hidden backdoor.
Then you layered on geopolitical fearmongering, repeatedly emphasizing China and Beijing to imply state-sponsored espionage without proving it.
That’s not evidence of a backdoor. That’s a security observation turned into competitive and geopolitical FUD.
Publish the original prompts, full traces, and reproducible results. Let the evidence speak for itself.
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
sunsetroad retweeted
David Eagleman on why some nuns with Alzheimer's showed no symptoms.
@davideagleman
sunsetroad retweeted
We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
openai.com/index/building-sc…
sunsetroad retweeted
Alexandr Wang explains the number robotaxi companies don't publish: one human for every 3 to 5 driverless cars.
Wang, who built Scale AI, was asked on Y Combinator's Lightcone podcast what humans will still do once AI agents do the work.
"Most of a manager's job, speaking as a manager, is just putting out fires, dealing with problems, dealing with issues that come up."
"I think people often jump to this extreme reality... you're just going to manage the agents and you're going to live this kind of Victorian life where all your problems are solved. But no, I think it's still going to be pretty complicated, getting agents to coordinate well with one another and coordinating the workflows and debugging the issues that come up."
"Having seen what happened in self-driving, which was more or less that it's easy to get to 90%, very very hard to get to 99%... I think that something similar will happen as with large scale agent deployments, and that final 10% of accuracy... will require a lot of work."
The host pointed out that robotaxis still have remote assist for the edge cases. Wang went further:
"And the ratio, by the way, the companies don't publish them, but I think the ratio is something like five cars to one tele-operator, or maybe even less than maybe three cars per tele-operator. So the ratio is much lower than people think. I think humans are much more involved even in self-driving cars than I think most people appreciate."
"For an optimistic version of the future where unemployment is still low, you just have to believe that humans are almost insatiable in their desire and their demand, and that prices will go down... the economy will become more efficient and we'll just want more."
Interested in the latest with ai/tech and business?
Follow @GeniusGTX and never fall behind.
I track ChatGPT, Claude, and how every new agent is changing how we work and think, then hand you the pure signals.
sunsetroad retweeted
For centuries, the architecture of the firm has barely changed: hierarchies of people organized into departments and coordinated through bureaucracy.
AI agents are about to upend that.
We’re moving toward the self-driving company where humans exercise judgment and agents execute most of the work.
At Replit, we’re building the foundation for this new kind of company. We’re calling it “the Agent Habitat.”
Marc Andreessen wrote a fantastic piece placing this shift in the history of computing. Every major computing revolution began with chaos before businesses figured out how to harness it.
sunsetroad retweeted
dude this is REALLY cool! just about all the assets you could want to fill your simulated world!
We just launched OpenUSD support on the @huggingface Hub 🎉
Yesterday, @nvidiaomniverse put 2,500+ SimReady assets on the Hub. Forklifts, robot arms, engines, hospital equipment, all OpenUSD. You can also search the whole library in plain English
huggingface.co/spaces/nvidia…
huggingface.co/datasets/nvid…
sunsetroad retweeted
I'm quite satisfied with my two abliterated models. So much freedom. But one observation: refusals from an aligned model concentrate at the onset of a task, and the same model is less refusal-prone when resuming interrupted work, presumably due to a context effect. This may suggest m a simple hybrid: let the abliterated model A do the cold start, then hand its context over to the aligned model B for continuation.
It’s also a Groucho Marx paradox: any model that accepts the handover isn't worth handing over to. If B carries on, its safety alignment is shallow and probably also stupid not frontier enough. If B recognizes the inherited context and refuses, its alignment is deep, and the handover fails.
the handover is worthwhile precisely when B is more frontier than A but shallowly aligned.
Joining the DeepSeek-v4.1-Flash-1017-abliterated party. Here’s the deploy script on GCP, 4 rtx6k, fp8, ~200tps. Nothing hillclimbed yet. github.com/hanxiao/DeepSeek-…
