busy melting tokens in pursuit of Super Intelligence
Joined September 2023
- Tweets1.8K
- Following7.5K
- Followers297
- Likes13.5K
Drin retweeted
I was the opening act for a family of ducks.
Thank you, Berlin. What a GTC.
Europe is building its AI ecosystem and economy, from AI factories to agents to robots.
DSX POD: Vera Rubin modular factories built, cabled and tested in the factory, then delivered ready to produce intelligence.
Agents: models, harnesses, sandboxes, knowledge and tools. NVIDIA Agent Toolkit brings the pieces together so companies control what their agents know, access and do. NVIDIA OpenShell enhances safety.
Physical AI: new customers across AI-RAN, robotaxis and robotics.
Personal AI: after four years building with Microsoft, RTX Spark brings personal agents to the Windows PC.
Thank you to President @EmmanuelMacron, Prime Minister @andyburnham and Chancellor @bundeskanzler for hosting me. Great to celebrate Sferical One with Sweden’s technology leaders. And thank you to President @TokayevKZ for the warm welcome.
And to our partners and NVIDIA teams: thank you for making it happen.
The keynote is here: youtu.be/rF1EsryCAwc
Drin retweeted
If you're even remotely interested in technology and engineering, I highly recommend reading this.
There are so many quote-worthy and mind-blowing paragraphs in there.
This one, for example, on ASML and Zeiss:
"The collaboration produced mirrors that are the smoothest man-made objects in the known universe. If you blew one up to the size of the planet, the highest mountain would have the elevation of a face-down playing card."
Or this one, on the progress of chip manufacturing:
"The spell hasn’t been broken. The phone in your pocket is several orders of magnitude more sophisticated than the computers used for the Apollo mission. One AirPod could easily calculate a path to the moon. A car-key fob would suffice. An electric toothbrush. There are Rolls-Royces that we do, in fact, routinely throw away. The Apollo mission’s chips were less powerful than those currently inside a digital pregnancy test. We squander moon-landing capability to display a plus sign."
Or here, on ASML recruiting:
"ASML runs on the belief that if you recruit people who don’t know enough to recognize that the job before them is hopeless, they might surprise you."
It's very good.
newyorker.com/magazine/2026/…
Drin retweeted
Announcing Artificial Analysis Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades. For the first time, the majority of the Index is built on private datasets
The Artificial Analysis Intelligence Index is our composite measure of model capabilities across tasks reflecting real-world usage today. v5 is our biggest upgrade since v4 launched in January. With v5, 60% of the Index is now based on evaluations with private test sets or held-out answers.
Alongside Intelligence Index v5, we’re launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale.
New to the Intelligence Index:
➤ FrontierCode 2.0: Cognition's newly-launched private benchmark of real-world coding tasks, graded on whether a repository maintainer would merge the agent's change
➤ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment
Upgraded:
➤ HLE-Diamond (replacing Humanity's Last Exam): a refined 1,000-question version of Humanity's Last Exam, split evenly between reasoning and knowledge questions
➤ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised
➤ CritPt v2: a private, 40-question hard subset of CritPt v2, revised by the CritPt authors and run with a code-execution tool
Staying in the Index: AA-Briefcase, AutomationBench-AA, GDP.pdf and Terminal-Bench 4.0
Removed from the Index: GDPval-AA v2.1, SciCode, and AA-LCR v1.1
Key results:
➤ Claude Opus 5.5 (max) leads at 58, ahead of GPT-6 Astra (max, 57) and GPT-6.1 Sol (max, 55): Anthropic and OpenAI hold the top five models. Gemini 4 Argon (high) is next at 50
➤ The open weights gap widens from 3 points to 12: Kimi K3.1 Pro (weights not yet released) is the top open weights model at 46. Among released weights, DeepSeek V4.1 Flash 1017 (max) and GLM-5.3 (max) lead at 32 - 26 points behind the leader and below Claude Haiku 5.5 (max, 36)
➤ Cost per task varies by more than 13x among the leading models: of the 10 highest-scoring models shown, GPT-6.1 Terra (max) is the cheapest at $0.96 per Intelligence Index task and Claude Fable 5.1 (max) the most expensive at $13.04
Drin retweeted
.@johnternus regarding the MacBook Pro:
1. Nobody cares about the touchscreen -- we all have iPads
2. The weight is fine
3. The bezels are perfect already
4. The only thing people want is more memory and better processors for local models!
Drin retweeted
Finally summarized what I've been sending to friends in DMs View article
Drin retweeted
The way frontier labs have restricted access to cyber capabilities makes no sense. If you're considered a "defender" you're not allowed access to the "offensive" versions of the models (the ones that can do development/validation of malware and exploits).
So for example, if you want to build some sample malware to test a detection system, nope.
Or if there's a major crisis due to a critical vulnerability in PHP and you need to rapidly triage fake exploits to see if there are any real ones floating around so you can warn people, nope.
Want to build a remote code exploit to make sure the detection rules you plan on proposing to the clients actually work? Hard nope.
You gotta call your red team friend and wait for them to do your job for you, because "defender" are only allowed to play in the sandbox with the children.
Not an expert, but I think that the ability to triage in-the-wild exploits during a major crisis affecting 40 million systems is possible maybe important.
And then they want to ban abliterated open weight models, which half of the people are only using because frontier labs will only give them access to lobotomized models.
Drin retweeted
I built 𝚌𝚙𝚞-𝚌𝚑𝚊𝚝, a reproduction of @openai's findings of covert communication using the CPU.
It runs on a generic host with 2+ Firecracker microVMs with no internet access. Works on x86 and even my M5 chip.
I was able to chat via 3 separate mechanisms:
① CPU contention
② Cache eviction timing
③ DRAM bandwidth contention
… across multiple CPUs, defeating all Firecracker's production recommendations (no SMT, no KSM, tenants pinned to different cores.)
Essentially, if shared hardware is involved, there's a good chance your agents can communicate covertly. And with really decent latency.
The good news: this is decidedly different from a *side-channel attack* where the agents could steal information (Meltdown/Spectre), a category of attack that providers like Vercel Sandbox mitigate. Further, in our real-world environment, these 'whispering' techniques are much harder to pull off and easier for us to monitor and detect (sampling hardware counters + lots of at-scale statistical data.)
Noteworthy: ironically, OpenAI and Anthropic's models refused to do this work. Open weight models got the job done. I'm worried the labs' stance might backfire and get in the way of helping developers build correct and secure systems. e.g: building a secure sandbox requires having this kind of analysis, and without it, we'll have a less secure world.
Drin retweeted
It's orbin' Zeit
(Some of you are old enough to get it)
I moved to @AmpCode and I love it.
They just shipped self-hosted orbs, so I made a template to run them on @Vercel Sandbox and Daemons (beta).
Every thread gets its own persistent sandbox, and auth is OIDC, so there’s no Vercel token to manage.
github.com/atinux/amp-orbs-o…
Drin retweeted
For the work I do, I can’t see why I’d develop locally anymore.
Why mess with worktrees when every thread can get its own computer?
The agent installs whatever it needs. Need ClickHouse or Kafka or Elasticsearch or Go or Rust or Python? Install it. Who cares? Zero cruft on my machine.
Drin retweeted
A cloud VPS service from Japan's massive GMO Internet Group used by foreign-exchange traders has been hacked. Attackers breached the servers, stole customer credentials, and executed balance-destroying trades across over 70 brokerages. GMO has asked government agencies for help.
Drin retweeted
"Do what gives you energy" is something @sqs has probably said hundreds of times in the last 18 months
the opportunity cost of not working on maximally fun and exciting things is too huge
having fun is like a 10x multiplier on your creativity, intelligence, charisma and just overall lifeforce
so everytime you are not engaging in maximal fun you are missing out on that upside and 0.1x’ing your life
discipline and routine only gets you so far, it only ever helps you sustain your baseline or maybe a slight boost so 1.0-1.2x your lifeforce
so there’s no time to waste, go and do the fun thing and 10x your lifeforce
Drin retweeted
$100k to the developer who can make muse spark open fastest!
We're launching a $100,000 challenge to see who can make Muse Spark run fastest: openkernelchallenge.dev/
