Links for 2026-03-24

Coding Agents Build Chess Engines in LaTeX and Learn Esoteric Languages from Scratch

Mathieu Acher, a computer science professor at INSA Rennes, has been conducting experiments that suggest recent coding agents are beginning to look less like autocomplete systems and more like adaptable problem-solvers: systems that can enter unfamiliar computational worlds, infer the rules, and build substantial working software with only minimal human guidance.

He asked Claude Code (Opus 4.6) and Codex CLI (GPT-5.2) to build chess engines from scratch across 12 languages, including ones where no chess engine has ever existed: LaTeX, Brainfuck, Rocq. No architecture docs, no step-by-step guidance. The results range from ~900 to ~2200 Elo depending on language. The TeX engine alone is ~2,100 lines of pure macro expansion, using \count registers as RAM and \csname tables as ROM (there’s nothing like it to copy from).

His latest experiment pushes the idea further. He challenged Claude Code with MNM Lang, an esoteric language invented in March 2026, in which source code is represented as colored M&Ms arranged on a grid. With essentially no ecosystem beyond the language spec and opcode table, the agent solved all 26 challenges, culminating in a working Brainfuck interpreter written in MNM Lang.

The key distinction, in Acher’s telling, is that these are not bare LLM completions but agentic systems that read specs, form hypotheses, run code, inspect failures, debug, and iterate. His broader claim is that the agent loop gives LLMs a real capacity to adapt to novel computational environments rather than merely reproducing familiar patterns from training data.

Sources:

1. Coding Agents Build Chess Engines From Scratch in Rust, C++, COBOL, Rocq, LaTeX, Brainfuck, and More https://blog.mathieuacher.com/FromScratchChessEnginesPolyglot/

2. TeXCCChess: How Coding Agents Wrote a Chess Engine in Pure TeX https://blog.mathieuacher.com/TeXCCChessEngine/

3. Can Coding Agents Program in M&Ms Language https://blog.mathieuacher.com/CodingAgentsMnMLang/

4. BFChess: A Chess Engine in Brainfuck, Built by a Coding Agent https://blog.mathieuacher.com/BFChessChessEngineBrainfuck/

See also:

https://x.com/ChaseBrowe32432/status/2035547317055668599

Hyperagents

Hyperagents: an AI system that not only improves at solving tasks, but also improves how it improves itself.

Hyperagents are self-referential agents that combine a task agent and a meta agent in one editable program, allowing them to modify both how they solve tasks and how they generate future improvements. The authors call this metacognitive self-modification: learning not just to perform better, but to improve at improving.

The main idea is to go beyond earlier recursive-improvement systems (Darwin Gödel Machine), which could improve task performance but still relied on a largely fixed, handcrafted self-improvement procedure. In DGM-H, the paper’s implementation, that meta-level procedure becomes editable too.

The most interesting result is evidence that the system can discover general self-improvement strategies. One example is the autonomous invention of persistent memory: instead of merely logging scores, the hyperagent stores synthesized insights, causal hypotheses, and forward-looking plans, then consults them during later self-modification steps. This lets later generations build on earlier discoveries and avoid repeating mistakes.

The paper also reports suggestive evidence of compounding self-improvements: improvements discovered in one run can be transferred to a new setting and continue accumulating.

Paper: https://arxiv.org/abs/2603.19461

Code: https://github.com/facebookresearch/Hyperagents

AI

1. Official confirmation that GPT-5.4 Pro was able to solve an open research math problem that humans were unable to solve in several attempts https://epochai.substack.com/p/first-ai-solution-on-frontiermath

2. GPT-5.4 Pro probably fully resolved a Machine Learning adjacent Open Math problem https://solveall.org/problem/gaussian-correlation-inequality-extensions

3. Siddharth Mishra-Sharma on how he got Claude to build a JAX-based cosmological solver in a few days. Similar efforts by domain experts have typically taken months to years. Claude compressed this into days. https://www.anthropic.com/research/long-running-Claude

4. AI Can Learn Scientific Taste https://www.alphaxiv.org/abs/2603.14473

5. The Y-Combinator for LLMs: Solving Long-Context Rot with λ-Calculus https://arxiv.org/abs/2603.20105

6. The Internalization of Computation – What Percepta’s transformer-computer actually signals — and why the implications are far stranger than the demo https://www.novaspivack.com/business/the-internalization-of-computation-what-perceptas-transformer-computer-actually-signals-and-why-the-implications-are-far-stranger-than-the-demo

7. China declares AGI development to be a part of 5-year plan https://www.lesswrong.com/posts/JKhoyebpedLQLb5JN/china-declares-agi-development-to-be-a-part-of-5-year-plan

8. Jensen answering a question from Lex on when he thinks an AI will be able to start, grow, and run a successful technology company worth more than a billion dollars: ‘I think it’s now. I think we have achieved AGI. ... It is not out of the question that a Claw was able to create a web service, some interesting little app that all of a sudden, you know, a few billion people used for 50 cents, and then it went out of business again shortly after. Now, we saw a whole bunch of those type of companies during the internet era, and most of those websites were not anything more sophisticated than what OpenClaw could generate today.’ https://www.youtube.com/watch?v=vif8NQcjVf0

9. OpenAl is offering private-equity firms a guaranteed minimum return of 17.5%, as well as early access to models not yet in public release. https://www.reuters.com/business/openai-sweetens-private-equity-pitch-amid-enterprise-turf-war-with-anthropic-2026-03-23/ [no paywall: https://archive.is/3MXXE]

10. Tesla, SpaceX Plan to Build New Chip Factory in Texas https://www.wsj.com/tech/tesla-spacex-plan-to-build-new-chip-factory-in-texas-4dfb9a46 [no paywall: https://archive.is/14Xgm]

11. New computer chip material inspired by the human brain could slash AI energy use https://www.cam.ac.uk/research/news/new-computer-chip-material-inspired-by-the-human-brain-could-slash-ai-energy-use

https://x.com/nasqret/status/2036213885426942423

Miscellaneous

1. Functional recovery of the adult murine hippocampus after cryopreservation by vitrification https://pubmed.ncbi.nlm.nih.gov/41774797/

2. “We map a range of systemic risks that could disrupt food production and distribution at scale. Building the resilience of food systems to withstand extreme shocks is a practical, actionable step that governments and institutions can take today.” https://allfed.info/research/publications-and-reports/peer-reviewed/its-time-to-consider-global-catastrophic-food-failures

3. Free math book: An Introduction to the Theory of Numbers http://www.trillia.com/moser-number.html

4. BYD just unveiled an electric car that can charge from 10 to 70 percent in five minutes, and all the way in nine https://www.wired.com/story/how-byds-ev-charger-got-even-faster-and-it-might-not-matter-as-much-as-you-think/ [no paywall: https://archive.is/oDlII]

5. Traders placed $580mn in oil bets ahead of Donald Trump’s social media post on Iran talks https://www.ft.com/content/1171d623-3709-4f6e-8ded-a5df4ec57696?syn-25a6b1a6=1 [no paywall: https://archive.is/KpdEN]

Ukraine

With no means to extinguish such massive blazes, for the second day in a row, Russians watch their country’s largest oil export facility burn on the Gulf of Finland.

Cloud cover usually prevents satellites from seeing strike damage, but not in Primorsk, Russia, where fires at the port are still burning for a second day.

1. From FP-1 to FP-7: How Ukraine’s Drone and Missile Program Went Ballistic https://united24media.com/war-in-ukraine/from-fp-1-to-fp-7-how-ukraines-drone-and-missile-program-went-ballistic-16841

2. Ukraine’s top drone commander wants to bleed Russia’s army dry https://www.economist.com/europe/2026/03/22/ukraines-top-drone-commander-wants-to-bleed-russias-army-dry [no paywall: https://archive.is/4UDZp]

3. The Russian army has sustained over 6,000 casualties in the last four days as it attempted a renewed offensive that was beaten back by the Ukrainian military. https://www.politico.eu/article/russian-troops-increased-pressure-on-ukraine-lost-over-6k-soldiers-in-4-days-kyiv-says/

Hungary

Orbán’s opponent, Péter Magyar, holds essentially the same political positions on issues such as immigration as Orbán. Magyar, a former Orbán supporter, left Orbán’s sphere due to widespread corruption within the Orbán system.

The most important difference between them is that Magyar is not a Russian asset. There is no good reason to support Orbán over Magyar other than wishing to help Russia achieve victory.

Foreign politicians who support Orbán are enemies of Europe. They are at least as bad as Angela Merkel, who caused irreparable damage to Germany.

See also:

1. How Viktor Orbán’s oligarchs reaped billions in public contracts https://www.ft.com/content/70c20ff6-9778-4f97-902b-d3c859b59339?syn-25a6b1a6=1 [no paywall: https://archive.is/0yQJs]

2. EU cuts Hungary out of sensitive talks over leaking-to-Russia fears, diplomats say https://www.politico.eu/article/eu-hungary-leak-russia-peter-szijjarto/