FeedCity

CBA has been naughty, and too reliant on their AI boyfriend. Here’s what happened: A CBA customer contacted bank requesting contact details for Secretlab (a company that is not us, and makes chairs) CBA staff member queried ChatGPT (possibly via their own personal, unauthen...

Reddit will block the Internet Archive

Well this sucks. Jay Peters for the Verge:

Reddit says that it has caught AI companies scraping its data from the Internet Archive’s Wayback Machine, so it’s going to start blocking the Internet Archive from indexing the vast majority of Reddit. The Wayback Machine will no longer be able to crawl post detail pages, comments, or profiles; instead, it will only be able to index the Reddit.com homepage, which effectively means Internet Archive will only be able to archive insights into which news headlines and posts were most popular on a given day.

Tags: internet-archive, reddit, scraping, ai, training-data, ai-ethics

Today's links Goodhart's Law (of AI): When a metric becomes a target, AI can hit it every time. Hey look at this: Delights to delectate. Object permanence: Bill Ayers graphic novel; Foxconn in India; Uber loses $4B; Warren Buffet, monopolist. Upcoming appearances: Where to fi...

If you've been experimenting with OpenAI's Codex CLI and have been frustrated that it's not possible to select text and copy it to the clipboard, at least when running in the Mac terminal (I genuinely didn't know it was possible to build a terminal app that disabled copy and...

qwen-image-mps Ivan Fioravanti built this Python CLI script for running the Qwen/Qwen-Image image generation model on an Apple silicon Mac, optionally using the Qwen-Image-Lightning LoRA to dramatically speed up generation. Ivan has tested it this on 512GB and 128GB machines...

AI for data engineers with Simon Willison I recorded an episode last week with Claire Giordano for the Talking Postgres podcast. The topic was "AI for data engineers" but we ended up covering an enjoyable range of different topics. How I got started programming with a Commo...

I’ve been reading listening to Poor Charlie’s Almanack which is a compilation of talks by Charlie Munger, legendary vice-chairman at Berkshire Hathaway. One thing Charlie talks about is what he calls “sit on your ass investing” which is the opposite of day trading. Rather th...

Chromium Docs: The Rule Of 2 Alex Russell pointed me to this principle in the Chromium security documentation as similar to my description of the lethal trifecta. First added in 2019, the Chromium guideline states: When you write code to parse, evaluate, or otherwise handle...

I've fallen a few days behind keeping up with Qwen. They released two new 4B models last week: Qwen3-4B-Instruct-2507 and its thinking equivalent Qwen3-4B-Thinking-2507. These are relatively tiny models that punch way above their weight. I’ve been running the 8bit GGUF vari...

I just gave my first solo main stage DEFCON talk, about Signalgate, and I think it was a resounding success! The room was packed – apparently people were waiting outside to get in. I got tons of positive feedback, and no negative feedback (though, someone told me a boomer sitting

say what you mean to say, not just the first thing on your mind

the percentage of users using reasoning models each day is significantly increasing; for example, for free users we went from <1% to 7%, and for plus users from 7% to 24%.

— Sam Altman, revealing quite how few people used the old model picker to upgrade from GPT-4o

Tags: openai, llm-reasoning, ai, llms, gpt-5, sam-altman, generative-ai, chatgpt

The Sound Blaster

Starting a blog about AI-assisted development, MCP integrations, and building production software with modern tooling like Claude Code and Cursor.

The issue with GPT-5 in a nutshell is that unless you pay for model switching & know to use GPT-5 Thinking or Pro, when you ask “GPT-5” you sometimes get the best available AI & sometimes get one of the worst AIs available and it might even switch within a single conversation.

— Ethan Mollick, highlighting that GPT-5 (high) ranks top on Artificial Analysis, GPT-5 (minimal) ranks lower than GPT-4.1

Tags: gpt-5, ethan-mollick, generative-ai, ai, llms

Intel released the 386 processor in 1985, the first 32-bit chip in the x86 line. This chip was packaged in a ceramic square with 132 gold-plated pins protruding from the underside, fitting into a socket on the motherboard. While this package may seem boring, a lot more is go...

Today's links Millionaire on billionaire violence: Let them fight. Hey look at this: Delights to delectate. Object permanence: Private equity vs investors; French teens who fought Nazis Upcoming appearances: Where to find me. Recent appearances: Where I've been. Latest books:...

You know what else we noticed in the interviews? Developers rarely mentioned “time saved” as the core benefit of working in this new way with agents. They were all about increasing ambition. We believe that means that we should update how we talk about (and measure) success when using these tools, and we should expect that after the initial efficiency gains our focus will be on raising the ceiling of the work and outcomes we can accomplish, which is a very different way of interpreting tool investments.

— Thomas Dohmke, CEO, GitHub

Tags: careers, coding-agents, ai-assisted-programming, generative-ai, ai, github, llms

When a Jira Ticket Can Steal Your Secrets Zenity Labs describe a classic lethal trifecta attack, this time against Cursor, MCP, Jira and Zendesk. They also have a short video demonstrating the issue. Zendesk support emails are often connected to Jira, such that incoming supp...

I gave a talk on Wednesday at the Bay Area AI Security Meetup about prompt injection, the lethal trifecta and the challenges of securing systems that use MCP. It wasn't recorded but I've created an annotated presentation with my slides and detailed notes on everything I talk...