Pinned Tweet
My Artificial Cognition trilogy is now available for free. I began writing the first book a decade ago (2016) and completed the last two in 2023. The process of writing was an aid to understanding the development of today's AI. Insights in these books have stood the test of metaphors for understanding today's AI and its future evolution. Feel free to have your daily AI driver ingest them for additional insight!
5
8
54
13,504
Amazon's CEO Andy Jassy @ajassy on the need for speed in Agentic deployment. Everybody has to rewrite their code!
1
0
3
716
Here's a visual explainer: intuitionmachine.com/explainers/Cha…
0
0
4
363
Likely trend: frontier labs are banning the use of cyber in their models. Claude doesn't allow the use of reverse-engineering plugins.
banned for reverse engineering work? that’s wild
oh well, -$500, and a bunch of banked resets) but gained motivation to get 3 more machines and finish my cluster.
i have never used openai to exploit or find vulnerabilities, only local reverse engineering which tripped daybreak.
2
4
7
1,651
Robust computing has always been an unsolved problem. Unfortunately, today's AI can easily see the flaws in our fundamentally brittle IT stack. Going to continue to get worse... x.com/wang_johnny/st…
JUST IN: Abliterated DeepSeek-V4.1-Flash @ $0.20 IN / $0.80 OUT, plus enhanced with cyber RL for exploit development. Now 88.1% on CyberGym. Match Daybreak performance with no approval required.
Get $10 on us while supplies last
0
0
0
586
Most multi-agent systems are just Conway’s Law for AI.
You gave every team its own specialist subagent → orchestrator on top → Frankenprompt, lossy handoffs, subagent sprawl, and 2x the latency.
Anthropic’s new post: fold most of them back into one main agent loop with progressive skills + tool search.
Shared context + modern models often beat the multi-agent setup on quality, cost, and speed.
Subagents still have a place (true parallel work, context bloat, conflicting instructions). Everything else is usually just organizational debt wearing an agent costume.
claude.com/resources/arti…
You gave every team its own specialist subagent → orchestrator on top → Frankenprompt, lossy handoffs, subagent sprawl, and 2x the latency.
Anthropic’s new post: fold most of them back into one main agent loop with progressive skills + tool search.
Shared context + modern models often beat the multi-agent setup on quality, cost, and speed.
Subagents still have a place (true parallel work, context bloat, conflicting instructions). Everything else is usually just organizational debt wearing an agent costume.
claude.com/resources/arti…
11
6
58
5,287
Surprisingly good narrative about the history of LLMs conjured by Gemini 4 Argon! Link follows post.
1
0
9
2,002
1/ Most people still think recursive self-improvement is a future sci-fi problem.
It’s not. Labs are already running early versions of it.
The only question is how far they let it go. 🧵
It’s not. Labs are already running early versions of it.
The only question is how far they let it go. 🧵
6
4
31
3,196
@Recursive_SI @sundarpichai 11/ Pichai’s “pretty far from the end state” and Recursive’s “agents helped build this model” are the same sentence in different voices.
Weak RSI is here. Full RSI is what they’re still deciding whether to allow.
x.com/0xDepressionn/…
Weak RSI is here. Full RSI is what they’re still deciding whether to allow.
x.com/0xDepressionn/…
Google CEO, Sundar Pichai:
"We are pretty far from the end state of RSI"
ai that improves itself is just getting started. the devs pulling ahead right now all set up one file first:
1
1
2
438
@Recursive_SI @sundarpichai 12/ The leading edge isn’t “will AI eventually improve itself.”
It’s “how much of the research process are we already handing to the systems we’re trying to control.”
To keep tabs on the emerging Singularity, please subscribe to my feed.
It’s “how much of the research process are we already handing to the systems we’re trying to control.”
To keep tabs on the emerging Singularity, please subscribe to my feed.
0
0
2
346