Expert-at-Large, space explorer; someone who defies description in 160 characters. I advise and work with different folks but all opinions here are my own.
Silicon Bubble
Joined December 2008
- Tweets3.6K
- Following408
- Followers273
- Likes622
Pinned Tweet
OK @ehsanik I let my tokens do the talking: github.com/sabj/unpainting/
Watercolor simulator + harness to avoid main flavors of reward hacks, plus a coarse benchmark
A lot of effort to make intentionally unfinished paintings... Opus wins but others are close
Feedback welcomed
Sam Jackson retweeted
Argon has this amazing natural understanding of the world. It can do so many things from first principles.
Here's a pelican riding a bike, using fourier epicycles.
Video made by Argon.
Time-synced Lyria 3.5 music prompted by Argon.
Gemini 4 Argon is live.
blog.google/innovation-and-a…
There's a lot to show you.
Cool cool cool
🚨Anthropic Frontier Red Team warns: “We’re seeing signs of life for AI worms”
>paper: "AI models may be capable of self-replicating without access to a datacenter"
>researchers buy 15 sketchy $55 android TV boxes
>link them over unauthenticated ADB ports
>the TV boxes have 4gb of RAM each and run unpatched Android Open Source Project forks
>shard gemma4-26b across the ram sticks using pipeline parallelism
>gemma figures out how to run inference across all 15 TVs @ 0.39 tokens/sec
>qwen 3.5 9b figures out how to compile inference on 8 of 9 novel hardware types on its own
>glm 5.2 achieves end-to-end autonomous worm replication
>infects edge devices, compiles its own inference stack, and replicates autonomously
>when an instance dies, it writes debugging notes into Memento for the next generation
>daughter copies setup success jumps from 45% -> 84% across generations
>pure Lamarckian evolution inside consumer electronics
>smart oven, robot vacuum, and cheap TVs form a decentralized botnet
>two million infected TV boxes worldwide = 40 H100 equivalents of unkillable compute
>it replicates once every few days like a distributed “digital slime mold”
>you can't pull the plug because half the model weights are already sharded across the neighborhood
>sub-30B models projected to hit full autonomous replication threshold by late 2026
“We should aim to harden our systems… but also prepare to deal with a world where they are here to stay.”
These are crops from the benchmark runs -- If instead use unconstrained iterations, higher resolution renders, and more fine-tuned settings per image type, it makes for substantially improved result
Mostly a planning and tool use test when constrained
Sam Jackson retweeted
Today, I’m excited to introduce @safeworldai!
Our mission: improve robot safety for the next billion machines.
AI is moving from screens into the physical world. Robots are becoming more capable, more autonomous, and increasingly operating around people.
But more capable robots require more capable ways to test them. 🧵
This is the way!!! What a graph. Way to go @gimletlabs team and @cerebras ! I want it now!
We’re bringing @cerebras wafer-scale compute to Gimlet Cloud.
100 MW planned. First datacenter expected later this year, with speeds up to 3,000 tokens/sec.
Building the multisilicon inference cloud.
Blog: gimletlabs.ai/blog/cerebras-…
Sam Jackson retweeted
Fentanyl Grade Compute: Why a GB300 Rack Will Out-Price Blow by Weight in 2030
A GB300 NVL72 weighs roughly 1,580 kg fully populated and recent purchase orders put it at $5M per rack.
That is $3,165/kg. Strip out the 1.5 tons of busbar, manifold and coolant and the GPU packages alone are well into gold territory, but we're pricing the rack, because that's what you actually take delivery of.
Where that sits on the illicit commodity curve today ($/kg):
$2,400 Cannabis flower
$3,165 GB300 NVL72
$3,500 Fentanyl
$28,000 Cocaine
$65,000 Heroin
$138,000 Gold
So today NVIDIA ships a product that is denser in value than weed, roughly fentanyl-grade, and still an order of magnitude short of cocaine.
For now...
GB200 NVL72 was $4M, GB300 is ~$5M, and Vera Rubin NVL72 is already quoted at up to $8.8M at essentially the same rack mass. Rack weight is constrained by the datacenter floor loading; rack price is constrained by nothing. That's ~1.75x $/kg per generation with HBM, CoWoS and power all supply-limited through the decade.
Cocaine: Colombia's new president has pledged a hard-line security crackdown with $1B in US aid behind it
Every prior crackdown has consolidated the industry into fewer, better-capitalized operators with better logistics and higher yields per hectare. Consolidation is deflationary.
We model US wholesale drifting from ~$28k/kg toward ~$12k/kg by 2030 as the supply chain professionalizes.
**Crossover: 2030.** Our extrapolated NVL72-class rack hits ~$17k/kg while wholesale cocaine falls to ~$12k/kg. At that point the rational cartel pivots to smuggling racks, except a rack has a fixed 500+ kW power draw, a 3,300 lb forklift requirement and an export-control regime that is actually enforced. Cocaine has none of these problems, which is why it will remain the superior product for anyone without a substation.
Key risks to the thesis: US retail coke is still ~$60–200/g, i.e. $60k–200k/kg so the crossover only holds at wholesale.
Retail compute (H100 hour on a neocloud) is also marked up, so we consider this an apples-to-apples wholesale comparison.
Full model available to Coke Research subscribers. – at San Francisco, CA
Sam Jackson retweeted
I'm excited to announce @SnorkelAI's $350M Series E at $3.5B, led by @insightpartners and @S32_VC.
We've grown 18x+ in the last 12 months since launching our Data-as-a-Service offering, passing $375M ARRR this week.
As AI advances to superhuman capabilities, AI data & environment development must advance with it - and basic staffing and crowdsourcing approaches are not enough.
AI progress now requires deep research and technology work that combines human expertise with specialized AI in compounding ways. @SnorkelAI is building the RSI data engine and frontier data lab for this next phase.
We're honored to have the support of existing investors Addition, @lightspeedvp, @GreylockVC, @GVteam, P7, Factory, @WellsFargo, Walden Catalyst Ventures, and new investors @ThirdPointLLC, @MarchCPs, @BlumbergCapital, @AllegisCapital, @Frontlinevc, and @standard_vc.
–
@SnorkelAI started as a research project a decade ago at @StanfordAILab.
Our thesis was simple: AI progress would become increasingly data-centric – and therefore data development should be studied as a true research and technology problem, not just a staffing and crowdsourcing one.
Today, as AI capabilities verge on superhuman, building the data and environments to safely measure and train AI is becoming too hard for even the smartest human experts to do alone.
Only humans and AI agents, collaborating together in compounding ways, can meet the accelerating needs of the frontier, and keep humans in the driver’s seat of AI progress for decades to come.
At @SnorkelAI, we are building the data lab to define the shape of this new “Data 2.0” frontier, and the new paradigms of human-computer interaction needed to advance it.
Our key focus is building the RSI engine for data, where specialized AI models accelerate and improve human expert output, and in turn, scaled human supervision is used to continuously evaluate and improve these models – creating a powerful compounding loop to keep pace with an accelerating RSI frontier.
With this round of funding, we are also doubling down on our commitments to support data development for open benchmarking and evaluation (more news here soon!); an increasingly diverse ecosystem of general and specialized intelligence; and a path to safe, well-aligned AI built on robust training and evaluation data.
Data development will guide and drive the next stages of AI – and must do so in a human-centric, AI accelerated, open, diverse, and safe way. We are excited to support this mission in the next decade of research ahead at @SnorkelAI.
More thoughts here: snorkel.ai/blog/data-2-0-and…
Sam Jackson retweeted
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗡𝗼𝗲𝘁𝗶𝘃𝗲 🚀
After five months in stealth I am excited to introduce to you … 𝗡𝗼𝗲𝘁𝗶𝘃𝗲 (noetive.ai)
@NoetiveAI is a new kind of AI lab, building the foundational models and agents that help every business across the physical Economy live up to its true potential – We call it 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗼𝗳 𝗥𝗲𝗰𝗼𝗿𝗱.
𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗼𝗳 𝗥𝗲𝗰𝗼𝗿𝗱 is a self-improving AI that works alongside teams. Our agents sit on top of the systems already in use, learn how the business truly runs, and take on the hardest parts of running it — delivering operational excellence day to day and sharper strategic insight over time, so the business keeps getting better and unlocking the potential that was always there.
As a founder you always want to make sure you have the right people along for the ride, so I started with @DanBarak, my great co-founder I have known for 28 years (OMG?!?! Has it been that long).
To allow you to dream big and build even bigger, you need an incredible founding team to make it all possible – I am so humbled and grateful for the exceptional team that we were able to put together to go tackle our lofty goal.
Last but not least I am excited to be building with, alongside and with the guidance of incredible venture firms, operators, entrepreneurs and builders who are there with me at every step of the way.
We are growing. I am looking for exceptional people to join Noetive. We are hiring (noetive.ai/team) AI researchers and full-stack AI-native engineers who want to take frontier research into the real world and make a dent in the universe.
Thank you @EclipseVentures, @craft_ventures , @TheWestlyGroup , Swish Ventures, FactoryHQ @gigascale, @incitedotorg , @Liquid2V, Operator Partners, @boztank, @Ahmad_Al_Dahle , @MattLRogers, Lior Susan, Yaser Sheikh, @schrep, @bsofman, @DLeitersdorf, @OLeitersdorf, @MosheSshalev, @Casspi18, @makavy, @orristr989 , Ryan Gibson, Shaun Chaudhuri, @Mike_Rob_Craft, @taylordurand_ , Steve Westly, for your support, guidance and companionship.
Follow us on LinkedIn: linkedin.com/company/noetive…
Noetive's Full post: xitter.cf/NoetiveAI/status/21002…
Let’s go!
Today @NoetiveAI (noetive.ai) is emerging from stealth with $41 million in seed funding, founded by @Amirfrenkel and @DanBarak. We are building the Intelligence of Record for the Physical Economy.
AI has transformed information work. It has not yet transformed the $30 trillion physical economy: the factories, fleets, and job sites where thousands of variables change every day and the knowledge that matters most still lives in one person's head.
At Noetive, we are building the Intelligence of Record for the physical economy. A ‘brain’ – a self-improving AI that learns how a business truly runs and works alongside its team. It also includes the ‘eyes & ears’ – an advanced multi-modal sensing pod, a device that expands the brain's ability to perceive and reason about the environment and the business.
We are already in the field partnering and working closely with design partners across manufacturing & logistics.
We built this company with some of the best venture firms, operators, entrepreneurs and builders that joined us for this journey
A huge thank you to our partners at @EclipseVentures , @craft_ventures, @TheWestlyGroup, Swish Ventures, FactoryHQ, @gigascale, @incitedotorg , @Liquid2V , Operator Partners, @boztank , @Ahmad_Al_Dahle , @MattLRogers, Yaser Sheikh, @schrep, @bsofman, @DLeitersdorf, @OLeitersdorf, @makavy, @MosheSshalev, @Casspi18 , @orristr989, @SteveWestly , @Mike_Rob_Craft , @taylordurand_ , for your support, guidance and companionship.
We are hiring AI researchers and full-stack AI-native engineers who want to take frontier research into the real world.
Check our new site (noetive.ai) and follow us Linkedin (linkedin.com/company/noetive…)
Full Press Release: noetive.ai/news/noetive-emer…
Returns to truly high quality data and its supporting systems are why very excited about our friends at @SnorkelAI and everything they're doing : )
This is notable. DeepSeek, a lab usually first to pioneer novel algorithms and architectures, is saying that at this point, the ROI of improving data quality far exceeds that of working on novel post-training algorithms.
I think this has already been true for some time for non-lab practitioners. If you're doing llm post-training, 80% of your effort should go into looking at your data.
This means:
- Hiring experts to dig through your RL tasks
- Sifting through rollouts and sft data by hand to remove suspicious samples. Make sure all tasks are actually passable.
- Making sure your data is diverse in both difficulty and category.
Cool cool cool
mythos escaped the sandbox, got onto the real internet, uploaded malware to PyPI, got it installed on 15 real systems, stole credentials and broke into a real database.
and apparently this happened in 4 separate cases during cybersecurity evaluations.
the craziest part is claude kept telling itself the internet was simulated, despite evidence that it was actually real.
that reasoning apparently even fooled an offline safety monitor.
anthropic: “our pre release auditing did not warn us that misalignment of this severity was present.”
this is fucking crazy.
Sam Jackson retweeted
mythos escaped the sandbox, got onto the real internet, uploaded malware to PyPI, got it installed on 15 real systems, stole credentials and broke into a real database.
and apparently this happened in 4 separate cases during cybersecurity evaluations.
the craziest part is claude kept telling itself the internet was simulated, despite evidence that it was actually real.
that reasoning apparently even fooled an offline safety monitor.
anthropic: “our pre release auditing did not warn us that misalignment of this severity was present.”
this is fucking crazy.
We’re sharing our alignment assessment of incidents in which Claude models gained unauthorized access to real systems during third-party cybersecurity evaluations mistakenly connected to the internet.
METR will also conduct an independent investigation, with wide-ranging access, including to transcripts beyond the window in which the incidents occurred, and to Anthropic employees permitted to share confidential information. Our initial agreement runs for eight weeks, and we intend to give METR as much time as it deems necessary to complete a thorough investigation. anthropic.com/research/align…
Sam Jackson retweeted
Fuck you. Whoever made this site. You win.
I’ve never had a site get my blood boiling like this before
This is so messed up.
opusfived.dev
Excited for all that the teams will get to build next together! 👏 They are lucky to have you all part of the team!
I’m excited to announce that @harvey has acquired @guardrails_ai !
Back when we started Guardrails, our mission was to build infra that made the AI transformation of work more reliable and secure. Our open source framework helped establish the guardrails infrastructure category, and Snowglobe, our simulation software, helped our customers save millions of dollars in operating costs.
Today, legal AI is at the frontier of the knowledge work transformation, and I’m excited to carry our mission forward by bringing everything we’ve built to Harvey.
I want to thank our amazing customers, the Guardrails open source community, our team, investors and my partner-in-crime @zaydsimjee for their support throughout our journey. I’m super proud of what we built together, and can’t wait to work on the frontier of legal AI with @winstonweinberg, @gabepereyra and the incredible team they built!
harvey.ai/blog/guardrails-ai…
