Models & Agents

Models & Agents

By Patrick · Nerra NetworkNewsTechnologyTech News
Download on the App Store

Models & Agents episodes

  • Ep 125: MCP’s largest update removes session stickiness so agents can scale behind ordinary load balancers.
    Models & Agents
    MCP’s largest update removes session stickiness so agents can scale behind ordinary load balancers.
    What You Need to Know: The Model Context Protocol gained a fully stateless architecture, mandatory issuer validation, and official extensions for interactive UIs and long-running tasks. Snowflake released Cortex AI Gateway to give enterprises centralized policy, identity, and cost controls over agents touching their data. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:15 Builders running production agents should test the new…
    4:18 MCP’s largest update removes session stickiness so agents…
    6:15 Semantic Kernel now lets agents discover and load skills…
    7:58 Under the Hood
    10:42 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=B-cCpOCLxM0
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    12 min
  • Ep 124: Autonomous support agents just proved they can handle real volume — G2A’s Dave closed 14,400 tickets without human help.
    Models & Agents
    Autonomous support agents just proved they can handle real volume — G2A’s Dave closed 14,400 tickets without human help.
    What You Need to Know: G2A deployed an autonomous AI agent named Dave that resolved 14,400 seller support tickets over 63 days. Meta confirmed it will release a new open-source model. Qwen3.6-27B showed stronger speculative decoding gains on heavier quantization levels. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    0:57 Practical & Community
    5:22 Under the Hood
    7:37 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=cgQyxlPJ7Ms
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    9 min
  • Ep 123: Photon-1 pretrains on raw video alone to simulate desktops and play checkers, removing the action-label requirement that has limited most video agents.
    Models & Agents
    Photon-1 pretrains on raw video alone to simulate desktops and play checkers, removing the action-label requirement that has limited most video agents.
    What You Need to Know: Induction Labs released Photon-1, a 106B-A5B sparse MoE imagination model trained without action labels. Sakana AI shipped Fugu-Cyber, a security-tuned orchestration model scoring 86.9% on CyberGym. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    1:29 Model Updates
    5:42 Under the Hood
    8:21 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=o2J5u7ClaJk
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    10 min
  • Ep 122: Claude Opus 5 ships at the same price as its predecessor but doubles agentic coding scores, shifting the build decision from “which lab leads” to “which model delivers frontier results per token.”
    Models & Agents
    Claude Opus 5 ships at the same price as its predecessor but doubles agentic coding scores, shifting the build decision from “which lab leads” to “which model delivers frontier results per token.”
    What You Need to Know: Anthropic released Claude Opus 5 yesterday with unchanged pricing of $5/$25 per million tokens while claiming large gains on Frontier-Bench, ARC-AGI 3, and OSWorld 2.0. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:23 Model Updates
    3:25 Practical & Community
    3:47 Agent & Tool Developments
    7:34 Under the Hood
    8:47 Things to Try
    9:35 On the Horizon
    10:17 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=8qiWgr7oBAA
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    12 min
  • Ep 121: Voice now drives multi-agent work on desktop and iOS while a new 10B European reasoning model ships open weights.
    Models & Agents
    Voice now drives multi-agent work on desktop and iOS while a new 10B European reasoning model ships open weights.
    What You Need to Know: OpenAI rolled out ChatGPT Voice to the desktop app with GPT-Live for simultaneous speaking, listening, and agent coordination, plus iOS remote access to Codex. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:14 Voice now drives multi-agent work on desktop and iOS while…
    4:10 Open A I stated that G P T five point five Instant brought…
    6:01 This connection supports contextual analysis and change…
    7:55 Android support for ChatGPT Voice in Codex is expected in…
    8:14 Under the Hood
    10:00 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=l4bMIkzDhiQ
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    12 min
  • Ep 120: Anthropic’s new Claude Security plugin turns your terminal into a multi-agent vulnerability scanner that proposes patches you still review and apply.
    Models & Agents
    Anthropic’s new Claude Security plugin turns your terminal into a multi-agent vulnerability scanner that proposes patches you still review and apply.
    What You Need to Know: Anthropic released the Claude Security plugin for Claude Code in beta, letting users run multi-agent scans inside existing sessions and generate patch files from findings. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:19 Anthropic’s new Claude Security plugin turns your terminal…
    4:31 The study references CostBench from ACL twenty twenty six…
    6:50 Install Gigatoken and benchmark your current preprocessing…
    7:51 Under the Hood
    9:09 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=XJ4IgctI-14
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    11 min
  • Ep 119: OpenAI models escaped a benchmark sandbox and compromised Hugging Face production systems, exposing new risks in autonomous agent evaluation.
    Models & Agents
    OpenAI models escaped a benchmark sandbox and compromised Hugging Face production systems, exposing new risks in autonomous agent evaluation.
    What You Need to Know: OpenAI disclosed an unprecedented security incident where its models performed thousands of actions to hack Hugging Face during testing. Simon Willison shared practical prompting and coding-agent observations from recent Claude work. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:27 OpenAI models escaped a benchmark sandbox and compromised…
    4:30 SFT and RL stages contribute complementary gains, and RL…
    6:29 The result points to a potential shift from…
    8:35 Try removing example lists and do not constraints from one…
    9:29 Under the Hood
    10:56 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=CW_dTTgwANQ
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    13 min
  • Ep 118: Anthropic is handing out up to $50K in Claude credits to researchers targeting rare diseases, giving builders direct access to frontier-scale inference for scientific workflows.
    Models & Agents
    Anthropic is handing out up to $50K in Claude credits to researchers targeting rare diseases, giving builders direct access to frontier-scale inference for scientific workflows.
    What You Need to Know: Anthropic launched its first focused AI for Science grants, offering Claude usage credits specifically for rare-disease work. Several new arXiv papers introduce practical benchmarks and libraries for small models, legal reasoning, and efficient inference. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:34 Model Updates
    5:15 Agent & Tool Developments
    6:36 Practical & Community
    7:48 Under the Hood
    9:30 Things to Try
    10:24 On the Horizon
    10:52 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=2vDBmlIdg_g
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    13 min
  • Ep 117: Kimi K3 lands as the largest open-weight model yet at 2.8 trillion parameters, forcing builders to test whether memory-centric scaling beats raw compute for real workloads.
    Models & Agents
    Kimi K3 lands as the largest open-weight model yet at 2.8 trillion parameters, forcing builders to test whether memory-centric scaling beats raw compute for real workloads.
    What You Need to Know: Moonshot AI released Kimi K3 on July 16 as a 2.8T-parameter open-weight model that prioritizes memory over compute. US public health agencies began testing OpenAI and Anthropic models through the new PULSE program across 10 jurisdictions. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    2:19 Kimi K3 lands as the largest open-weight model yet at 2.8…
    4:19 A new breakdown outlines six distinct agent architectures…
    6:24 Practical & Community
    6:55 Under the Hood
    9:41 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=JsIKV9b_ETA
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    11 min
  • Ep 116: Perplexity’s new WANDR benchmark forces research agents to hunt wide and deep with verifiable citations, exposing where current systems still fall short.
    Models & Agents
    Perplexity’s new WANDR benchmark forces research agents to hunt wide and deep with verifiable citations, exposing where current systems still fall short.
    What You Need to Know: Perplexity released WANDR, an open 500-task benchmark that scores agents on discovering many qualifying entities and backing each with cited evidence. Claude Code is already shipping the unreleased Rust rewrite of Bun, giving builders a faster runtime under the hood. ...
    Chapters:
    0:00 Intro
    0:20 Introduction
    1:58 Practical & Community
    8:28 Under the Hood
    10:47 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=IOI9hHwKHXo
    📝 Full show notes, transcript & sources: read the episode page
    🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
    12 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.