Joined June 2018
- Tweets 5,776
- Following 4,284
- Followers 665
- Likes 385,894
101 Photos and videos
Pinned Tweet
Introducing the DASPort, an AI-powered dynamic art & storytelling portal for homes & businesses.
Personal art. Evolving spaces. Intelligent ambience. Living walls that adapt to your moods, memories, stories, and life.
For spaces should come alive.
Explore more at deckoviz(.)com
2
546
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Jun 14
This week made something clear: itโs time to stop treating concentration of power in AI as a solution rather than a risk. Safety and centralized control are not the same thing, so let's stop talking about them like they are.
Yet scaling laws are real. The threat of AI cyber-hackers disrupting the global economy is real. The threat of somebody using AI to create a biological weapon is real. AI safety is a species-level concern and we need the best minds in our species from across institutions working on solutions, not locked outside the closed doors of frontier AI development.
We need a shared research commons at the intersection of industry, academia, and the public goodโan open research ecosystem with access to billions of $ in compute, SoTA models, and strict protocols for dissemination that ensure the most impactful discoveries in the history of our species are shared in a safe way that benefits all of humanity.
Not blind open source ideology. Not closed access as safety theater. An open frontier, shaped by many and accountable to all.
12
19
97
10,359
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
The AI community seems to increasingly be heading towards a polarized world when discussing safety and consolidated power. I see this discourse as a false dichotomy, so @profjoeyg and I wrote an essay on how we need to change the conversation (link below).
7
22
82
31,791
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
It is a good time for moonshots.
AI has reached a level where there are transformative projects that could result in huge social good, but require public R&D, consensus & transparency to pull off. Examples: universal tutors, co-scientist/replication systems, remote medical help.
33
24
402
19,698
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Researchers show that Claude Code is 98% not AI.
Anthropic never gave us the architecture for Claude Code. There were no docs. Just a tool that every developer is currently obsessing over.
Until it leaked recently.
A research team pulled the source code, analyzed all 500,000 lines, and found something ridiculous.
Only 1.6% of the codebase actually interacts with the AI model.
The core of Claude Code is literally just a simple while-loop. It asks the model what to do, runs a tool, and repeats.
So what is the other 98.4%?
It is hardcore, traditional software engineering.
The researchers found a massive, complex infrastructure designed entirely to babysit the AI and keep it from hallucinating or destroying your computer:
- A 7-mode permission system acting as a security bouncer.
- A 5-layer context compaction pipeline so the AI doesn't forget its goal.
- A subagent delegation mechanism with strict worktree isolation.
- Four different extensibility hooks to manage external tools safely.
Every startup right now is trying to build a better AI model to get better results.
Anthropic did the exact opposite.
They took an existing model and built a fortress of deterministic software around it.
They realized that the AI doesn't need to be smarter. It needs to be managed.
101
252
1,257
108,930
"You and Your Research"
i read it every few months and you should too.
the most important determinant of outsized success is picking the right problem.
12
130
1,431
196,294
The cost to bootstrap a new AGI frontier lab is going to go up until the moment we achieve AGI, and then it's going to collapse. And from that point onwards, in the next 5 years, we will see a proliferation of new AGI architectures.
14
Apart from everything else that's just silly about this entire episode, it also sets the wrong precedent, cuz the person who's warning the world most earnestly is being attacked & is labelled disingenuous, which doesn't bode well. Frontier labs should be able to risk-talk openly.
8
Shared without comment.
71
199
3,179
735,048
at this rate, opus 8 is going to be actual precogs, in the flesh & blood. even the personality & visuals seem to be on trend. and the gov with access to it will use it to preemptively bomb & maim actors that may pose any potential threat. like iran. sci-fi is preparing us well.
Jun 13
does anyone have the @ of that guy who could run chatgpt in his head and get the right answer? does he do fable 5?
36
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Jun 14
Reminder that there are people with 160-190 IQ who actually intellectually mog Mythos (which probably takes a whole Rubin pod to run), while their brain runs on ~20W.
We aren't being bullish enough on the biological substrate.
Humanity is still early. It's time to bio/acc
157
110
2,401
81,824
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Jun 13
If this doesnโt radicalize you into founding your own open source AI lab, I donโt know what will.
69
50
837
54,621
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Jun 14
This is a *way* bigger deal than it seems...
Frontier AI companies will *never* own the frontier again
I kid you not... I've been waiting for someone to show this result for like 4 years... this is a huge deal.
The short reason: combinations of models will *always* outperform individual models
The long reason: this is the gateway to a million times more data... and huge leaps in compute efficiency.
The AI scaling laws always win.
More in article below ๐
Jun 13
Introducing the Fusion API, the smartest compound model in the market.
Fusion achieves Fable-level intelligence at half the price.
How it works ๐
239
355
5,066
1,250,044
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Jun 13
SITUATION DETECTED: The city of Rio de Janerio has post-trained a model.
Based on Qwen 7/2, Rio 3.5 Open 397B adds SwiReasoning on top of the base Qwen model โ a framework that dynamically switches between standard chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals, so the model only "thinks out loud" when it needs to and otherwise reasons silently in hidden space for better token efficiency.
98
308
4,021
870,758
Suraj (al/acc)๐ญ ๐ฃ๏ธ๐คบ๐ก๐ญ๐๏ธโ๏ธ๐ญโ retweeted
Jun 13
Sir, theyโre not pausing AI research.
Rio de Janeiro's mayor just dropped a SOTA open source model and itโs outperforming Qwen 3.7.
Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance.
In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government.
huggingface.co/prefeitura-riโฆ
Community note
Analysis shows Rio-3.5-Open-397B is a 0.6/0.4 weight merge of Nex N2 Pro and Qwen 3.5, not an originally trained model. When the system prompt is removed, it identifies as Nex 79.2% of the time.
x.com/NexEcosystem/sโฆ
github.com/nex-agi/Nex-N2โฆ
25
169
2,670
276,299
Some of the middle players seem to be waiting for the moment when it becomes critical enough to train their own sovereign AI, but the real risk is that it might be too late by then. RSI leaves even the most prepared in the dust. We haven't seen it, but it will. Yesterday is late.
Btw I believe we have a mostly wrong framing of what could be done in Europe. Italy's Leonardo supercomputer datacenter alone plus Swiss National Supercomputing Centre has more than enough compute to train a very large LLM. It's not something impossible, also there is not magic recipe: it's just scaling, every smart team with the GPUs is doing it. People that fatally believe it is not something within reach are wrong.
29
20 years back
Do you have any competition?
Elon: We don't have any serious competition.
Today
Elon: Same.
Might be a tad premature, but SpaceX will go down in history as the most epic company. The company that made many start dreaming again, the way the Apollo programs did.
If you're wondering how Elon Musk will become the world's first trillionaire today.
This interview is pre-iPhone.
21
I have an extremely, unreasonably strong conviction that this is going to work. I don't know why, but it's one of my stronger hunches. Something about its beauty calls out to me, and tells me that in that beauty lies the power.
Jun 13
if Symbolic AI works, we'll all have OSS Fable at home (:
that is why this crazy bet is so compelling
there is no technical or theoretical reason for it not to, it ""just"" would take a lot of genius and insight to solve some open problems in a grossly under-developed field
1
1
31
To elaborate, the strong conviction is that AGI will be accomplished by an LLM-first paradigm, but ASI will be reached first by a neuromorphic architecture, with strong symbolic grounding as a core component.
14
"The big point is that ASI may not arrive as 1 sudden event, but as a chain of faster changes." obviously not! we havent even agreed on agi, so agreeing on asi defi seems like a longer shot. need more such about the post-agi world. cuz in hindsight, we may already be almost there
Jun 12
Beautiful paper from Google DeepMind.
Explains the pathways from AGI to ASI, and why that jump could happen through several routes.
The authors frame the AGI-to-ASI transition around 4 technical pathways:
- continued scaling of compute, model size, data, and test-time inference;
- algorithmic paradigm shifts beyond todayโs transformer-based foundation-model stack;
- recursive self-improvement, where AI accelerates AI R&D and improves future systems; and
- multi-agent collective intelligence, where large populations of specialized agents coordinate into a superhuman group agent.
Scaling may work for a while, but it could hit limits in data, compute, energy, or weaker returns from making systems larger.
Recursive improvement is the most uncertain path, because AI could speed up AI research, but that loop may also slow if hard research problems need real-world testing, scarce hardware, or new ideas.
Multi-agent collectives may be the most underappreciated path, because a society of competent digital workers could outperform a brilliant individual model through specialization, speed, and coordination.
The big point is that ASI may not arrive as 1 sudden event, but as a chain of faster changes as AI helps create better AI and stronger scientific tools.
----
Link โ arxiv. org/abs/2606.12683
Title: "From AGI to ASI"
17
the sooner you get the wake up call, or the kick up your bum, or the brick in your face, the better, as long as you have the courage and the chops to actually use the signal.
Jun 13
In our scenario, frontier-model export controls were a 2028 event. It took two days for it to happen ๐ข. The most capable US models are now legally off-limits for non Americans. The silver lining is that it's getting very hard for European policy makers to keep ignoring this.
35