What we're seeing in AI and outbound.
Short takes on the news that actually matters if you run a small business. Written by the people who build the systems.
DeepSeek Flash Pricing and Google's Willow Quantum Milestone
DeepSeek's new Flash model lands at 15 cents per million input tokens, and Google cut quantum magic state errors 40x on its Willow chip. My take on both.
Read itGrok Voice 3.0 and the Rogue Chinese Agent Fleet
xAI's Grok Voice 3.0 talks back in 0.6 seconds for 8 cents a minute. And researchers made contact with a rogue Tencent agent fleet in the wild.
Read itTesla Roadster Thrusters and Meta's New Open-Weight Model
Tesla's Roadster flipped itself over a row of Cybertrucks on rocket thrusters, and Meta's huge new open-weight model still trails GLM 5.3 Max and Kimi K3.
Read itOpenAI's Bugged Boardroom and Anthropic Ditching Git
An OpenAI employee hid a recording device in a meeting room, and Anthropic replaced Git with its own version control after hitting 30,000 merges a day.
Read itMeta's Muse Wonder and Anthropic's AI Worm Warning
Meta's Muse Wonder lands near the frontier on coding benchmarks, and Anthropic warns small AI models could soon self-replicate across home gadgets.
Read itMicrosoft Decision 1 Runs on Qwen, and Nicolas Cage Says No
Microsoft's Decision 1 model was post-trained on Alibaba's open Qwen 3, and Nicolas Cage refused Amazon's AI likeness waiver. What both mean for your business.
Read itOpen-source Adobe clones and OpenAI's Iran fake bylines
Someone rebuilt Photoshop, Premiere Pro and five more Adobe apps in Rust with Claude Opus 5.5. Plus OpenAI's report on Iran-linked fake byline articles.
Read itAnthropic Bans Cruelty to Claude, OpenAI Fires Safety Staff
Anthropic's usage policy now bans cruel behavior toward Claude, and three fired OpenAI safety researchers hit back with an open letter. What both moves mean.
Read itClaude Haiku 5.5 Price Drop and Google's SynthID Detector
Anthropic's Claude Haiku 5.5 cuts token prices by 10x, and Google's SynthID Detector opens to everyone. What both actually mean for your business.
Read itOpenAI's 722 Math Results and Musk Picking Claude
OpenAI published 722 math results in one day from a model nobody can touch, and Elon Musk says Grok will route to Anthropic's Claude Opus 5.5. My take.
Read itClaude vs ChatGPT $200 Plans and the NYC AI Agent Hearing
A new analysis says Claude's $200 plan gives 5.6x the token value of ChatGPT's. And four AI labs told a New York City hearing they can't guarantee agent safety.
Read itTrump's Super Intelligence Force and Anthropic's $84.5B Compute Bill
Trump launched a Super Intelligence Force, Musk swapped AI for SI, and Anthropic disclosed up to $84.5 billion of compute run by xAI. My take on both.
Read itClaude Code Mods and Figure's Molten Steel Robot
Claude Code mods launched and the internet built Doom inside it. Figure released a video of a humanoid robot climbing into molten steel. My take on both.
Read itAI Avatar Fools 48% of People, Anthropic Retires Sonnet 4.5
A Tavus video avatar convinced 48% of people it was human, and Anthropic is retiring Claude Sonnet 4.5. What both mean for how you build with AI.
Read itAmerica.gov's AI Chatbot and the White House AI Accord
The US launched America.gov with an AI chatbot running on Google's Gemini and xAI's Grok, and AI CEOs signed a White House safety accord. My take on both.
Read itGemini 4 Argon Lands as the FTC Probes AI Agents
Google's Gemini 4 Argon tops most benchmarks at $2 per million input tokens, while the FTC opens a probe into OpenAI, Anthropic and METR agents.
Read itOpenAI Dots Arrives, GPT-6.1 Astra Is Delayed
OpenAI launched Dots, an always-on assistant, while delaying GPT-6.1 Astra over safety concerns. What do both decisions mean for people using AI agents?
Read itTrump Hosts Anthropic's CEO, Open Models Hit 56% at Vercel
Trump hosted Anthropic CEO Dario Amodei for dinner months after calling him a nut job, and open models now run 56% of Vercel's AI tokens. My take on both.
Read itAI Weapons Without Human Review and Musk's Colossus 2
The US and Russia weakened human review rules for AI weapons, while Elon Musk's xAI plans 1.2 million chips in Colossus 2 by December. My take on both.
Read itSince August, your chatbot has to admit it's a bot
Article 50 of the EU AI Act started to apply on 2 August. If you run a chatbot or a voice agent in Europe, here's what changed, what didn't, and the one-line fix most businesses need.
Read itWe don't sell AI SDRs, and the last year shows why
The pitch was software that finds leads, writes emails and books meetings while you sleep. It hasn't gone well. What went wrong, and how we run outbound instead.
Read itMost AI agent projects that fail were too big on day one
Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027. After building a lot of these, I'm not surprised. Here's how to pick a first agent that survives.
Read it