←Search results

Search “Agents”

By relevance · Total 428 results Clear search

Friday

  1. 36Kr · AISelected

    DeepMind AlphaProof Nexus appears in《Science》: autonomously solves 9 Erdős problems

    Google DeepMind's automated mathematical proof system AlphaProof Nexus appeared in the October 8 issue of《Science》, autonomously solving 9 of 353 open Erdős problems, two of which had remained open since 1970. The system uses Gemini 3.1 Pro to draft proofs in Lean, with a compiler verifying line by line and feeding back iterations; the reasoning cost per problem is several hundred dollars. The paper's ablation results show that a simplified system that removes complex structures such as the sub-agent tournament and retains only the large model plus Lean feedback can also solve these 9 problems, though the hardest few are more expensive.

    3 open Erdős problems, two of which had remained open since 1970. The system uses Gemini 3.1 Pro to draft proofs in Lean, with a compiler verifying line by line and feeding back iterations; the reasoning cost per problem is several hundred dollars. The paper's ablation results show that a simplified system that removes complex structures such as the sub-agent tournament and retains only the large model plus Lean feedback can also solve these 9 problems, though the hardest few are more expensive.

    Matched fieldsBody

    #hot-signal#Papers/Research#Reasoning#DeepMind
  5. Simon Willison

    sparqlite: Implementing SPARQL 1.1 inside SQLite in Rust

    Simon Willison used Claude Code's Ultracode mode to generate sparqlite, a Rust-written loadable SQLite extension that stores RDF quads in ordinary SQLite tables and executes SPARQL 1.1 queries and updates inside the SQLite process. By integer-encoding terms, using covering indexes, a cost-based planner, batched joins, and optional direct reads of SQLite B-trees, it is 3.5 times faster than SQL translation in count mode and 2.8 times faster when returning full results, and it passes 636 of the 639 applicable W3C evaluation tests. The author says the experiment kept Claude Code running continuously for more than 12 hours; he cannot review the resulting code and is unsure how to use this implementation built on a whim.

    Simon Willison used Claude Code's Ultracode mode to generate sparqlite, a Rust-written loadable SQLite extension that stores RDF quads in ordinary SQLite tables and executes SPARQL 1.1 queries and updates inside the SQLite process. By integer-encoding terms, using covering indexes, a cost-based planner, batched joins, and optional direct reads of SQLite B-trees, it is 3.5 times faster than SQL translation in count mode and 2.8 times faster when returning full results, and it passes 636 of the 63

    Matched fieldsBody

    #newsletter#English text is being prepared.#English text is being prepared.#Agents

Thursday

Wednesday

  9. The New Stack

    Anthropic red team tests AI self-replication with 15 cheap TV boxes

    Anthropic's Frontier Red Team built a cluster of 15 off-brand Android TV boxes for about $825, ran AI models at 0.39 tokens/second, and assessed the feasibility of small-model self-replication in a paper. The experiment showed that Qwen3.5-9B could deploy a working Llama3.1-8B inference service on 8 of 9 hardware types, and GLM-5.2 covered all 8 simulated nodes in 4 of 5 trials. Using the Epoch Capabilities Index, the researchers projected that by the end of 2026 the strongest open-source models under 30 billion parameters will reach GLM-5.2's general capabilities, but stressed that the experiment was in a controlled environment and left conveniences in place.

    Anthropic's Frontier Red Team built a cluster of 15 off-brand Android TV boxes for about $825, ran AI models at 0.39 tokens/second, and assessed the feasibility of small-model self-replication in a paper. The experiment showed that Qwen3.5-9B could deploy a working Llama3.1-8B inference service on 8 of 9 hardware types, and GLM-5.2 covered all 8 simulated nodes in 4 of 5 trials. Using the Epoch Capabilities Index, the researchers projected that by the end of 2026 the strongest open-source models

    Matched fieldsBody

    #hot-signal#Papers/Research#Safety / alignment#Agents