Was reading a Hacker News discussion today where one person brought up one of the most hilarious, simple and effective mass law enforcement actions ever: Operation Flagship.
https://en.wikipedia.org/wiki/Operation_Flagship
"Operation Flagship was a sting operation jointly organized by the United States Marshals Service and the Metropolitan Police Department in Washington, D.C. that resulted in the arrest of 101 wanted fugitives on December 15, 1985."
"The fugitives voluntarily went to the Washington Convention Center, responding to an invitation sent by the fictitious television company Flagship International Sports Television, (which had the same initials, F.I.S.T., as Fugitive Investigative Strike Team) to claim two free tickets to watch the Washington Redskins American football home game against the Cincinnati Bengals and for a chance to win tickets to Super Bowl XX. A total of 166 marshals and police officers were involved in the operation, with undercover personnel posing as tuxedo-wearing ushers, cheerleaders, emcees, caterers, mascots, and maintenance staff."
They could do this all day long, and probably could create something similar to ensnare a whole bunch of English-speaking cybercriminals. Like tell them all there's a fancy watch convention that is giving away a whole crate of Trezor wallets made of gold.
And it's not just Cursor... rival agentic coding IDE Windsurf announced their own custom RL-trained fast coding model today as well!
Here are notes and a pelican on Windsurf's new SWE-1.5 model https://simonwillison.net/2025/Oct/29/swe-15/
The other day we had our first ever chained AI tool success on the #curl factory floor:
- tool A found a possible flaw in code and reported it.
- using the plain English description from tool A, tool B could create a reproducible by itself that verified the finding
The sense of magic is strong in this.
Now us poor humans need to fix it. The AIs are still really lousy at writing patches.
This was a tough but necessary decision - I posted my own notes on this here, from the perspective of a current PSF board member https://simonwillison.net/2025/Oct/27/psf-withdrawn-proposal/
https://fosstodon.org/@ThePSF/115446659188615376
Juicy little tidbit, this result holds when controlling for masculinity insecurity! "prototypicality legitimacy" is its own distinct belief system here (likely domain specific imho)
Source: Danbold, F., & Huo, Y. J. (2017). Men's defense of their prototypicality undermines the success of women in STEM initiatives. Journal of Experimental Social Psychology, 72, 57-66.
I found a copy here: https://huolab.psych.ucla.edu/wp-content/uploads/sites/5/2013/03/Danbold-Huo-JESP-2017.pdf
earlier today i reproduced running an amdgpu (polaris10) as eGPU on MNT Reform with RK3588 over PCIe breakout, something i didn't expect to become possible. i also got a glimpse of sway working (instead of just kmscube) but my cables are too wonky for a stable connection. after all the main quests i have to do atm i'm looking forward to come back to this, as it has a lot of potential for running more high powered apps and games with our systems
I gave a talk last night about "Living dangerously with Claude", on the joys and perils of --dangerously-skip-permissions and how critical it is that we run coding agents in a sandbox so that we can unlock their full potential https://simonwillison.net/2025/Oct/22/living-dangerously-with-claude/
First look at the DGX Spark
My notes on Claude Code for web, Anthropic's new asynchronous coding agent - I had preview access over the weekend, it's effectively a sandboxed instance of "claude --dangerously-skip-permissions" running in Anthropic's container https://simonwillison.net/2025/Oct/20/claude-code-for-web/
Traditionally, financial crashes are the best time to get tighter regulations passed, so now is probably a good time to start planning for the AI crash. Here is the starting point for my wishlist:
Companies may not issue dividends or share buybacks while they have any outstanding debt.
Companies above a $1bn valuation may not invest in companies up or down their supply chain without regulatory approval.
Companies above a $500bn valuation may not invest in companies up or down their supply chain at all.
Companies with a valuation above $500bn are automatically investigated for anticompetitive behaviour every three years. If they have not addressed all of the issues found in one investigation by the time the next takes place, they are split up into multiple companies, with no exceptions.
The rules for when companies must be taken public (and therefore have public reporting duties) are adjusted to apply to beneficial owners, having 500 people invest in a fund that invests in the company counts as 500 investors, not one.
Any CxO found to have misled investors is stripped of all shares they were awarded, any income from sale of shares, and any earnings above the median wage of their company, and banned from ever holding company directorships or CxO positions again.
What’s on your list?
Welcome to Day 2 of Dr. Cat's Mastodon Lecture Series: Underrated Women at the Intersection of Humanity * Engineering
Have you ever said to yourself my god if only someone could come along and stop reducing software development work to stupid metrics! If only they would sit next to some real working teams and just LISTEN to us!
You may be looking for: MARIAN PETRE
Because we need good news: Gary Larson is drawing again (he bought a tablet, taught himself how to use it, and suddenly drawing was fun again): https://www.thefarside.com/new-stuff/
NVIDIA sent me preview hardware of their new DGX Spark 128GB ARM64 4TB "AI supercomputer" - it's a very neat little device, here are my notes so far
https://simonwillison.net/2025/Oct/14/nvidia-dgx-spark/
nanochat by Andrej Karpathy is neat - 8,000 lines of code (mostly Python, a tiny bit of Rust) that can train an LLM on $100 of rented cloud compute which can then be served with a web chat UI on a much smaller machine https://simonwillison.net/2025/Oct/13/nanochat/
code / data wrangler in Switzerland.
Recovering reply guy. Posts random photos once in a while.