Hacker News: hardmaru

Sakana AI Launches Its First Commercial Product, Sakana Marlin

hardmaru — Mon, 15 Jun 2026 14:55:22 +0000

Article URL: https://sakana.ai/marlin-release/#English

Comments URL: https://news.ycombinator.com/item?id=48542201

Points: 2

# Comments: 0

Sakana AI's Recursive Self-Improvement (RSI) Lab

hardmaru — Fri, 05 Jun 2026 17:27:39 +0000

Article URL: https://sakana.ai/rsi-lab/

Comments URL: https://news.ycombinator.com/item?id=48415633

Points: 29

# Comments: 23

DiffusionBlocks: Training Neural Networks One Block at a Time

hardmaru — Thu, 28 May 2026 07:51:03 +0000

Article URL: https://pub.sakana.ai/diffusionblocks/

Comments URL: https://news.ycombinator.com/item?id=48305949

Points: 5

# Comments: 0

Reproducing all of Schmidhuber's papers with Claude

hardmaru — Fri, 08 May 2026 17:02:28 +0000

Article URL: https://cybertronai.github.io/schmidhuber-problems/visual-tour.html

Comments URL: https://news.ycombinator.com/item?id=48065843

Points: 4

# Comments: 0

Sparser, Faster, Lighter Transformer Language Models

hardmaru — Fri, 08 May 2026 16:44:21 +0000

Article URL: https://pub.sakana.ai/sparser-faster-llms/

Comments URL: https://news.ycombinator.com/item?id=48065594

Points: 4

# Comments: 0

New comment by hardmaru in "Apple already won the AI race: Your favorite AI will be gone soon"

hardmaru — Mon, 27 Apr 2026 06:50:36 +0000

No paywall link: https://archive.ph/HvOtP

Apple already won the AI race: Your favorite AI will be gone soon

hardmaru — Mon, 27 Apr 2026 06:50:09 +0000

Article URL: https://medium.com/@michalmalewicz/your-favorite-ai-will-be-gone-soon-9b3b035fe3db

Comments URL: https://news.ycombinator.com/item?id=47918469

Points: 3

# Comments: 2

Towards end-to-end automation of AI research

hardmaru — Fri, 24 Apr 2026 14:51:47 +0000

Article URL: https://www.nature.com/articles/s41586-026-10265-5

Comments URL: https://news.ycombinator.com/item?id=47891101

Points: 3

# Comments: 0

Sakana Fugu: A Multi-Agent Orchestration System as a Foundation Model

hardmaru — Fri, 24 Apr 2026 00:56:50 +0000

Article URL: https://sakana.ai/fugu-beta/

Comments URL: https://news.ycombinator.com/item?id=47884224

Points: 1

# Comments: 0

String Seed of Thought: Prompting for Distribution-Faithful, Diverse Generation

hardmaru — Tue, 21 Apr 2026 02:33:07 +0000

Article URL: https://pub.sakana.ai/ssot/

Comments URL: https://news.ycombinator.com/item?id=47843848

Points: 1

# Comments: 0

Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automata

hardmaru — Tue, 21 Apr 2026 02:17:14 +0000

Article URL: https://pub.sakana.ai/digital-ecosystem/

Comments URL: https://news.ycombinator.com/item?id=47843742

Points: 2

# Comments: 0

Can LLMs Flip Coins in Their Heads?

hardmaru — Mon, 20 Apr 2026 15:43:51 +0000

Article URL: https://pub.sakana.ai/ssot/

Comments URL: https://news.ycombinator.com/item?id=47835968

Points: 1

# Comments: 0

Towards end-to-end automation of AI research

hardmaru — Sun, 05 Apr 2026 02:48:00 +0000

Article URL: https://www.nature.com/articles/s41586-026-10265-5

Comments URL: https://news.ycombinator.com/item?id=47645696

Points: 6

# Comments: 0

Towards end-to-end automation of AI research

hardmaru — Wed, 25 Mar 2026 16:42:33 +0000

Article URL: https://www.nature.com/articles/s41586-026-10265-5

Comments URL: https://news.ycombinator.com/item?id=47519786

Points: 2

# Comments: 0

Sakana AI Announces Strategic Partnership with Google

hardmaru — Fri, 23 Jan 2026 13:22:49 +0000

Article URL: https://sakana.ai/google/#en

Comments URL: https://news.ycombinator.com/item?id=46732188

Points: 3

# Comments: 0

An Unofficial Guide to Prepare for a Research Position Application at Sakana AI

hardmaru — Tue, 20 Jan 2026 07:44:32 +0000

Article URL: https://pub.sakana.ai/Unofficial_Guide/

Comments URL: https://news.ycombinator.com/item?id=46689027

Points: 2

# Comments: 0

Repo: Language Models with Context Re-Positioning

hardmaru — Mon, 19 Jan 2026 00:49:15 +0000

Article URL: https://pub.sakana.ai/repo/

Comments URL: https://news.ycombinator.com/item?id=46673805

Points: 2

# Comments: 0

The Mythology of Conscious AI

hardmaru — Thu, 15 Jan 2026 04:01:56 +0000

Article URL: https://www.noemamag.com/the-mythology-of-conscious-ai/

Comments URL: https://news.ycombinator.com/item?id=46627851

Points: 6

# Comments: 7

DroPE: Extending the Context of LLMs by Dropping Their Positional Embeddings

hardmaru — Mon, 12 Jan 2026 06:36:27 +0000

Article URL: https://pub.sakana.ai/DroPE/

Comments URL: https://news.ycombinator.com/item?id=46584855

Points: 5

# Comments: 1

New comment by hardmaru in "Digital Red Queen: Adversarial Program Evolution in Core War with LLMs"

hardmaru — Thu, 08 Jan 2026 16:17:29 +0000

Hi HN,

I am one of the authors from Sakana AI and MIT. We just released this paper where we hooked up LLMs to the classic 1984 programming game Core War. For those who haven't played it, Core War involves writing assembly programs in a language called Redcode that battle for control of a virtual computer's memory. You win by crashing the opponent's process while keeping yours running. It is a Turing-complete environment where code and data share the same address space, which leads to some very chaotic self-modifying code dynamics.

We did not just ask the model to write winning code from scratch. Instead, we treated the LLM as a mutation operator within a quality-diversity algorithm called MAP-Elites. The system runs an adversarial evolutionary loop where new warriors are continually evolved to defeat the champions of all previous rounds. We call this Digital Red Queen because it mimics the biological hypothesis that species must continually adapt just to survive against changing competitors.

The most interesting result for us was observing convergent evolution. We ran independent experiments starting from completely different random seeds, yet the populations consistently gravitated toward similar behavioral phenotypes, specifically regarding memory coverage and thread spawning. It mirrors how biological species independently evolve similar traits like eyes to solve similar problems. We also found that this training loop produced generalist warriors that were robust even against human-written strategies they had never encountered during training.

We think Core War is an under-utilized sandbox for studying these kinds of adversarial dynamics. It lets us simulate how automated systems might eventually compete for computational resources in the real world, but in a totally isolated environment. The simulation code and the prompts we used are open source on GitHub.