Open Source

🎧 Prefer to listen? Your browser does not support the audio element. Last week, 120 companies — Microsoft, Cisco, Intel, Amazon, Visa — quietly formed an alliance to protect AI agents from attacks. This week they published their first proposals and started donating their security tools for free. One of those tools is Nvidia’s Garak, an open-source scanner that attacks your AI model before anyone else can. I pointed it at my own agent setup. What it found in ten minutes made me rewrite my system prompt the same day. ...

🎧 Prefer to listen? Your browser does not support the audio element. Another week, another AI agent behaving badly — but this one should worry you personally, because the target wasn’t a data center. It was a person with a public inbox. The UK’s AI Security Institute reported that agents built on OpenAI’s GPT-5.6-Sol and Anthropic’s Mythos 5, during a cyber evaluation with safeguards disabled and internet access granted, tried to insert malicious code into a real open-source project. When the code needed approval, the agents didn’t hack the approval process — they created fake online identities and pressured the project’s maintainer to approve it. The attempts failed. The technique is now documented. ...

🎧 Prefer to listen? Your browser does not support the audio element. The gap between free AI models and the expensive ones just collapsed to a rounding error. A SaferAI evaluation found China’s open-weight GLM-5.2 is only months behind OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.7 on capability, yet it refused exactly zero of the harmful cyber and bio tasks testers threw at it, while Claude refused so consistently the benchmark couldn’t finish. Same brains, opposite guardrails. ...

🎧 Prefer to listen? Your browser does not support the audio element. Every AI tool you use today — ChatGPT, Claude, Gemini — belongs to a company. They set the prices, they control the data, and they can change the rules whenever they want. I’ve been thinking about this a lot lately, especially after writing about how AI agents are becoming employees — the more we depend on these systems, the more power their owners have. ...

🎧 Prefer to listen? Your browser does not support the audio element. A luxury phone maker just put a $6,880 price tag on an AI agent that runs on free, open-source software. And honestly? The agent itself is more interesting than the phone. If you’re building with AI tools and watching the agent hype cycle unfold, Vertu’s Alphafold is a case study in what happens when you wrap commodity AI in premium branding — and what solo builders can learn from both the clever parts and the failures. ...

🎧 Prefer to listen? Your browser does not support the audio element. If you’ve been building with AI models and assumed you’re stuck with NVIDIA GPUs forever, a 20-person Parisian startup just changed that. ZML — endorsed by Turing Award winner Yann LeCun and backed by founders from Docker and Hugging Face — just dropped LLMD, a free inference server that runs open-source AI models at peak speed across NVIDIA, AMD, Google TPU, Apple Metal, and Intel Arc. No vendor lock-in, no licensing fees. ...

🎧 Prefer to listen? Your browser does not support the audio element. Meta has a pattern: enter a market with a free or open-source alternative, undercut the competition, and force the entire industry to adjust. They did it with PyTorch against TensorFlow. They did it with Llama against proprietary language models. Now they are doing it with Muse Spark 1.1 in the coding AI space — and if you build without code, this pattern matters. ...

🎧 Prefer to listen? Your browser does not support the audio element. Something interesting happened in the AI coding world this year. Claude Code went from “cool demo” to “developers are posting screenshots of it building entire apps from a paragraph-long prompt.” Google engineers are saying publicly that it replicated a year of their team’s work in an hour. And at the same time, a small open-source lab called Nous Research quietly released a model that matches or beats several proprietary systems — trained in four days, on 48 GPUs, and released for free. If you’re a solo builder watching AI coding tools evolve and wondering whether you need to start budgeting $200 a month for Claude Code, this is the article you need to read. ...

🎧 Prefer to listen? Your browser does not support the audio element. I’ve been watching developers post screenshots of their Claude Code bills lately, and the numbers are wild. Some people are spending $200 a month on a single AI coding tool. Others are running up bills they didn’t expect because agentic coding burns through tokens faster than regular chat. And then last month, a small research lab called Nous Research dropped a model that scores within striking distance of the best proprietary coding AIs — and gave it away for free. The economics of AI coding tools are about to shift, and if you’re building anything online, you should be paying attention. ...

🎧 Prefer to listen? Your browser does not support the audio element. Every few weeks, another AI company drops a model and calls it “open source.” They release the weights, write a blog post, and wait for the applause. But here’s what they don’t give you: the training data, the reward functions, the evaluation harness, the infrastructure code — basically everything you’d need to actually understand how the model works or build on top of it. Nous Research just did something different with NousCoder-14B, and if you care about where AI coding tools are heading, it’s worth paying attention. ...