ThinkingCap - The Local Coding Model

In this video, Sam Witteveen looks at ThinkingCap, a fine-tune of the Qwen3.6-27B model. BottleCap's goal was intentionally conservative: do not try to make the model smarter or teach it new capabilities. They wanted to preserve knowledge, reasoning ability, answer quality, conversational style, instruction following, and safety behavior.

The only thing they wanted to change was the amount of computation spent reaching an answer. In other words: keeping the same model, and making sure it overthinks less.

Anthropic Just Fixed Graph Engineering's Greatest Flaw

AI LABS explores the technical shift from loop engineering to graph engineering in AI workflows. This guide details how to implement robust verification strategies, skill orchestration, and browser-based testing to improve agent reliability and manage complex task outputs effectively.

Boris Cherny: We Cut 80% of Claude Code’s Prompt

Fresh off the launch of Opus 5, Claude Code creator Boris Cherny joins Diana Hu at Startup School 2026 to talk about what the newest models can do, how Claude Code came to be, and what it means to build products when the underlying capabilities keep accelerating.

MCP just got its biggest update ever — here’s what changes
for AI agents

The Model Context Protocol, the open standard that has quietly become the connective tissue between AI agents and the world's software, is getting its largest update since Anthropic released it twenty months ago — a sweeping architectural revision that its maintainers and backers say finally makes agentic AI ready for massive enterprise production deployments.

The update, released today under the stewardship of the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation, finalizes MCP's transition to a fully stateless architecture, hardens its authentication model against a known class of attacks, establishes a formal 12-month deprecation policy, and graduates two headline capabilities — interactive server-rendered interfaces and long-running asynchronous tasks — into official protocol extensions.

Anthropic launches Opus 5

On Friday, Anthropic launched its Opus 5 model, the newest version of its long-standing heavyweight model. While smaller than Fable 5, the model will be both cheaper and less restrictive than Fable, likely making it preferable in most use cases.

Opus 5 is launching only two months after Opus 4.8, which became available on May 28. Mythos 5, Fable 5, and Sonnet 5 all launched in June, leaving only the lightweight Haiku model still waiting for an upgrade to the 5 series.

Poolside drops Laguna S 2.1, an open-weight coding model
that beats rivals 10x its size

Poolside, the San Francisco AI lab that has spent most of its three-year existence quietly selling coding models to governments and defense agencies, released its most capable model to date — and made an unusually aggressive bet that radical transparency, not raw scale, is how a smaller lab competes at the frontier.

The model, Laguna S 2.1, is a 118-billion-parameter Mixture-of-Experts (MoE) system that activates only 8 billion parameters per token, supports a context window of up to 1 million tokens, and — according to benchmarks published by the company — matches or beats open models several times its size on agentic coding tasks. The weights are available immediately on Hugging Face under the permissive OpenMDW-1.1 license.

AMD Ryzen AI Halo - 100% Local AI

Sam Witteveen explores the capabilities of the new AMD Ryzen AI Halo, which features 128GB of unified memory. The demonstration highlights how this architecture enables local operation of large language models, image generation tools, and fine-tuning workflows, while comparing the practical benefits of unified memory against traditional workstation setups with discrete GPUs.

Google releases three new Gemini models — but no 3.5 Pro

Google DeepMind has released Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. Gemini 3.6 Flash is Google’s “workhorse model” that promises improved capabilities in coding, knowledge work, and multimodal performance while reducing token usage by up to 17%, making it cheaper than its predecessor 3.5 Flash.

Google is working on a new AI chip designed to make
Gemini more efficient

Alphabet, Google’s parent company, is designing a new server chip to help its in-house Gemini models operate more efficiently.

The new chip, internally dubbed “Frozen v2,” is slated to be released sometime in 2028, The Information reported, citing anonymous sources. According to the report, the chip could be between six and 10 times more efficient than Google’s existing AI chips, measured by the number of tokens generated per unit of power.

Pi is the Claude Code Killer Nobody Saw Coming...

Alan Walsh examines the critical role coding harnesses play in AI performance, comparing the feature-rich, "batteries-included" approach of tools like Claude Code with the modular, minimalist design of Pi. Learn how these competing philosophies influence agent behavior and discover strategies for building and extending custom coding environments tailored to specific development needs.

China’s Moonshot AI releases Kimi K3, the largest open-source model ever, rivaling top U.S. systems

Moonshot AI, the Beijing-based artificial intelligence startup backed by Alibaba, on Thursday released Kimi K3 — a 2.8-trillion-parameter model that the company says is now the largest open-source AI model in the world, and one that benchmarks show performs neck-and-neck with the most powerful proprietary systems from Anthropic and OpenAI.

The release, timed to land just ahead of the 2026 World Artificial Intelligence Conference in Shanghai, is a dramatic escalation in the global AI arms race and a watershed moment for the open-source AI movement. It also marks a remarkable comeback for a company whose market position had eroded significantly over the past 18 months following DeepSeek's meteoric rise.

Thinking Machines amps up its bet against one-size-fits-all AI
with its first open model, Inkling

Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, released its first in-house AI model Wednesday morning, called Inkling. And unlike the flagship models from OpenAI, Anthropic, or Google, it’s open-weight, meaning outside developers and companies can download it and modify it directly.

Inkling is a mixture-of-experts system with 975 billion total parameters, though it only draws on a fraction of that — about 41 billion — for any given task, a common design that keeps very large models faster and cheaper to run. It was trained on 45 trillion tokens of text, image, audio, and video, and reasons natively across all four.

How to Start Writing Loops for Advanced AI Models
like Fable 5 + GPT 5.6 (Clearly Explained)

Everyone is talking about writing loops instead of prompting agents like Claude Code, Codex and Cursor. But what actually are loops and how do you write them?

In this video I break down the agent loop in simple terms so that you can start writing loops that 10x your outcomes from agents like Claude Code, Codex and Cursor - even if you're non technical.

Because loops don't just have to be for software engineering, they can also work for marketing, design, research, strategy, data analysis and much more.

This New Google Format Gives Your AI Agent a Second Brain

A few weeks ago Google Cloud announced the Open Knowledge Format, or OKF — a vendor-neutral open standard that, in their words, "formalizes the LLM-wiki pattern" into a portable folder of markdown for handing an AI agent the curated context it needs.

I pointed Codex at my own full-stack app and asked it to walk the codebase and generate a full knowledge layer: one concept per markdown file, each one pointing at where the live code actually lives rather than copying it. Around 45 concept files, cross-linked into a browsable graph. 

GPT-5.6 IS HERE! BEST AI Model Ever? Beats Fable,
Faster, & Cheaper!

OpenAI has officially launched the GPT-5.6 family with Sol, Terra, and Luna, and after extensively testing every model, it's safe to say OpenAI is back in a big way.

Meta enters the crowded AI coding battle with Muse Spark 1.1

Meta publicly launched a new version of Muse Spark on Thursday, a multimodal AI model designed for agentic coding that aims to compete with similar products offered by OpenAI and Anthropic.

Spark 1.1, the first version of which was announced in April, can engage in multistep reasoning and handle complex processes, manage digital workflows, and deploy new features in enterprise systems.

OpenAI releases new voice models for more natural live conversations

OpenAI today released new conversational models, called GPT-Live-1 and GPT-Live-1 mini, claiming that they sound more natural and can handle turn-taking better. These are full-duplex models, meaning they can speak and listen at the same time, allowing users to interrupt naturally and enabling features like live translation.

How to build a custom AI harness with Claude SDK

Everybody is saying, “It’s not the model, it’s the harness,” but almost nobody stops to explain what a harness actually is. So I did. I built one live on the show: a Sentry bug-debugging harness for my company ChatPRD, using the Claude Agent SDK, a custom terminal UI built with the Ink library, and opinionated adapters for Sentry, Linear, GitHub, and Vercel.

I also walk through the architecture, share the code structure, and give you the exact process I used so you can build your own harness for any repetitive, structured workflow in your business.

DeepSeek open sources DSpark, a new framework to speed up
LLM inference by up to 85%

DeepSeek is back with yet another open release that could once again change AI development around the globe. 

Over the weekend, the firm released DSpark, a new, MIT-Licensed system designed to make large language models answer faster without changing what the underlying model is trying to say. 

The easiest way to think about it is this: most AI chatbots write like someone crossing a river one stepping stone at a time. They choose one small chunk of text, then the next, then the next. 

DSpark gives the system a scout that runs a few steps ahead, guesses the likely path, and lets the larger model quickly check which steps are safe. When the guesses are good, the model moves faster. When the guesses are weak, DSpark tries not to waste time checking them.

Herdr in about 6 minutes

If you run multiple coding agents like Claude Code, Codex, and Pi, herdr makes it much easier to manage everything in one terminal workspace. In this video I walk through installing herdr, setting up integrations, using workspaces/tabs/panes, tracking agent state, restoring sessions, and even spinning up sub-agents from inside Pi.

Z.ai launches ZCode to challenge Cursor, Claude Code
and GitHub Copilot in AI coding

Z.ai, the Beijing-based artificial intelligence lab formerly known as Zhipu AI, on Wednesday officially launched ZCode, a free desktop application it describes as an "Agentic Development Environment" purpose-built for its flagship GLM-5.2 large language model. The move marks the company's most aggressive push yet into the fast-growing AI-powered coding tool market, where it now competes directly with Cursor, Claude Code, GitHub Copilot, and Google's Antigravity.

Anthropic launches Claude Sonnet 5 as a cheaper way
to run agents

As shipping agentic capabilities becomes table stakes among foundation model companies, Anthropic is releasing Claude Sonnet 5, a more powerful and agentic version of the lab’s midsize model. 

Sonnet 5’s pitch is confirmation that agentic capability is the new baseline expectation at every price tier. Now the differentiator isn’t going to be who can do agentic work best, but how cheaply they can do it and how reliably without human oversight.