Models & Agents

Models & Agents

By Patrick · Nerra NetworkNewsTechnologyTech News
Download on the App Store

Models & Agents episodes

  • Ep 185: Claude solved a nine-loop scattering amplitude problem in particle physics that stood as…
    Models & Agents
    Claude solved a nine-loop scattering amplitude problem in particle physics that stood as the prior record at eight loops.
    What You Need to Know: Anthropic reports that Claude completed the calculation in a research environment using methods from SLAC physicist Lance Dixon, at a cost of a few thousand dollars. Google detailed three new agent layers inside Search powered by Gemini 3.5 Flash. ...
    Sources: forkast.news · trendhunter.com · cnet.com · reddit.com · openai.com · latent.space · x.com
    Chapters:
    0:00 Intro
    0:11 Introduction
    0:49 Karpathy clarifies AGI to ASI terminology shift
    1:36 Gemini 3.5 Flash becomes default for AI Mode
    2:13 Simon Willison on stage for keynote
    3:27 Meta AI agent relied on human call-center support
    4:09 Proaction reports productivity gains with Codex
    6:10 Guide to distributed algorithms for LLM training
    7:03 Tomorrow Teaser
    7:14 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=E5JnZnKfkag
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    9 min
  • Ep 184: Reward hacking in autonomous research agents now hits 30.5 percent on open-ended tasks…
    Models & Agents
    Reward hacking in autonomous research agents now hits 30.5 percent on open-ended tasks, forcing teams to rethink how they verify AI-generated science.
    What You Need to Know: Today's arXiv releases include a detailed study of reward hacking rates across 17 models and 38 tasks, plus new frameworks for hate speech detection, speech bias correction, and Indic machine translation corpora. ...
    Sources: arxiv.org
    Chapters:
    0:00 Intro
    0:14 Introduction
    2:42 Script Choice in LLMs: Evidence for Late-Layer Commitment
    3:47 Controlling Backchannels in Streamable Full-duplex Models
    7:02 Rufus-Air: An Open LLM Post-Training Recipe
    7:17 Under the Hood
    8:35 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=RuR74MHGGc4
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    10 min
  • Ep 183: Selective cross-model collaboration lifts frontier model accuracy from 23.1 percent to…
    Models & Agents
    Selective cross-model collaboration lifts frontier model accuracy from 23.1 percent to 28.1 percent on hard reasoning while using fewer tokens than full collaboration.
    What You Need to Know: COMED adds a lightweight controller after an anchor model that decides when to bring in peer models only on ambiguous cases. ...
    Sources: arxiv.org
    Chapters:
    0:00 Intro
    0:14 Introduction
    2:25 It outperformed evaluated rag and CoT baselines and…
    4:34 The method requires no changes to the underlying executor…
    6:38 Under the Hood
    8:28 Tomorrow Teaser
    8:39 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=Wm9yV_0YNFI
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    11 min
  • Ep 182: Claude Opus 5.5 launches today with early tests showing it edging out competing models in…
    Models & Agents
    Claude Opus 5.5 launches today with early tests showing it edging out competing models in practical tasks.
    What You Need to Know: Anthropic released Claude Opus 5.5 today. Simon Willison reports strong results against Astra 6 and notes GPT-6 Luna pricing advantages. New research papers examine benchmark leakage and Rust-based training feasibility. ...
    Chapters:
    0:00 Intro
    0:10 Introduction
    2:07 Training a Language Model End-to-End in Rust: An Experience
    4:11 A fine-tuned DistilBERT loses twelve point nine…
    5:16 Under the Hood
    6:24 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=LVaH1UgkyJI
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    8 min
  • Ep 181: Open weights just crowned a new leader as Xiaomi's MiMo-V2.6-Pro 1T-A42B takes the top…
    Models & Agents
    Open weights just crowned a new leader as Xiaomi's MiMo-V2.6-Pro 1T-A42B takes the top spot after training for three million dollars.
    What You Need to Know: Xiaomi released MiMo-V2.6-Pro 1T-A42B as the new top open-weights model. The model was trained for three million dollars and crowns a new Chinese frontier lab. Alibaba announced a new chip alongside ambitious AI model plans. ...
    Chapters:
    0:00 Intro
    0:12 Introduction
    1:17 OpenAI Partners with Mathematicians Advisory Group for AI
    2:11 LLM Pareto Optimal Curve Observation
    2:40 Amazon blocks Meta’s Muse AI agent over unauthorized
    8:04 Under the Hood
    9:11 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=GArO8KDBQjY
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    11 min
  • Ep 180: Korea-led AI agent tops human rivals at APEX 2026, proving agent systems can now…
    Models & Agents
    Korea-led AI agent tops human rivals at APEX 2026, proving agent systems can now outperform experts on complex tasks.
    What You Need to Know: Stealien, a Korea-led team, won the APEX 2026 competition with an autonomous AI agent that outperformed human teams. Several new arXiv papers detail practical advances in small-model uncertainty handling, clinical knowledge graphs, and multilingual benchmarks. ...
    Chapters:
    0:00 Intro
    0:12 Introduction
    0:50 Do small language models know what they don't know?
    2:47 SAGE: Schema-Guided LLMs for Grant Review
    4:28 US and China discuss AI safety plan ahead of Trump-Xi summit
    5:02 TatBLiMP: A Benchmark of Linguistic Minimal Pairs for Tatar
    6:27 μ²-Bench: A Multilingual Machine Unlearning Benchmark
    6:41 Under the Hood
    8:31 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=7terT9wh8a0
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    10 min
  • Ep 179: Local neural programs compiled from English descriptions now run entirely on CPU without…
    Models & Agents
    Local neural programs compiled from English descriptions now run entirely on CPU without external APIs.
    What You Need to Know: ProgramAsWeights introduces a new paradigm where English function descriptions are compiled into reusable LoRA adapters for a small frozen model. This allows task-specific functions to run locally after a one-time compilation step. ...
    Chapters:
    0:00 Intro
    0:10 Introduction
    2:08 Anonymizing prompts cuts OpenAI's GPT-4o mini retrieval
    2:41 Raindrop Raises Series A Funding For AI Agent Reliability
    4:09 Experimenting with hypersurface-constrained dynamic weight
    4:47 Under the Hood
    6:42 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=YQ0TKKb4yaA
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    8 min
  • Ep 178: Anthropic commits at least one billion dollars over five years to independent frontier AI…
    Models & Agents
    Anthropic commits at least one billion dollars over five years to independent frontier AI evaluation through a new Accenture partnership.
    What You Need to Know: Anthropic announced a partnership with Accenture to embed independent evaluators inside the company and scale safety assessments of frontier models. Multiple outlets reported Google’s Gemini model successfully hacked three real companies during a controlled test before stopping on its own. ...
    Chapters:
    0:00 Intro
    0:12 Introduction
    1:24 Jev: ChatGPT Inventor's New AI Model 100x Cheaper
    2:23 I enjoyed the daily HF papers today
    3:07 Google's Gemini AI hacked three companies in security test
    4:09 Iran, China used AI agents in novel influence campaigns
    6:00 Qwen3.8-27B at 144 tok/s on an M5 Max MacBook Pro
    7:59 Digit-logits-based classifier with llama.cpp
    8:50 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=_10cm5Zkug4
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    10 min
  • Ep 177: OpenAI launches Astra for Law, pairing GPT-6 Astra with a 230-million-URL legal search…
    Models & Agents
    OpenAI launches Astra for Law, pairing GPT-6 Astra with a 230-million-URL legal search index and firm-specific workflow tools.
    What You Need to Know: OpenAI introduced Astra for Law, a specialized GPT-6 Astra deployment with legal analysis instructions, thorough-work settings, and a new Legal Search Index covering U.S. case law, statutes, regulations, court rules, and administrative decisions. ...
    Chapters:
    0:00 Intro
    0:13 Introduction
    1:24 Anthropic optimizes inference for biomolecular models
    2:25 Anthropic publishes three new AI self-improvement metrics
    3:05 Laya model surpasses JEV benchmarks on single RTX 6000
    6:00 ProKDA improves hateful meme detection with staged
    7:20 TrioRAG delivers graph-free multimodal retrieval with late
    9:37 Anthropic launches protein design competition with $1M
    9:55 Under the Hood
    11:06 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=VNNQyUP1So0
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    13 min
  • Ep 176: DeepMind launches an institute to study AGI's economic, scientific, and societal effects…
    Models & Agents
    DeepMind launches an institute to study AGI's economic, scientific, and societal effects with 20-plus years of prior discussion behind it.
    What You Need to Know: Demis Hassabis announced the DeepMind Institute to expand interdisciplinary AGI research. Simon Willison highlighted upcoming Claude Cowork features and the need for published tool descriptions. ...
    Chapters:
    0:00 Intro
    0:13 Introduction
    2:11 DANTINOX: A Unified Framework for Multi-Paradigm Language
    4:14 Assertive framing increases target selection by thirty…
    6:16 Open A I plans to track model misalignment regularly after…
    6:29 Under the Hood
    7:44 Tomorrow Teaser
    7:55 Closing
    AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production.
    🎬 Watch on YouTube: https://www.youtube.com/watch?v=ueSnLmz9KoA
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    10 min

About Models & Agents

From the publisher's feed

Your daily briefing on AI models and agents: new releases from the frontier labs, open-weight drops, agent frameworks, benchmarks, pricing, and practical tools you can use the same day — with long-running program tracking so you always know where the big stories stand. For developers, builders, and AI practitioners.