2026-09-07
A Deception Probe Result Changed When I Averaged Different Response Tokens
SummaryIn a previous post, I trained linear probes on role-playing responses and tested them on sandbagging responses. Across the five models that I tested, the probes generally ranked deceptive sandbagging...
C64 - How to use the Scroller code in SEUCK Redux mode
How to assemble the Commodore 64 SEUCK Redux version of the Scroller engine source code. Review the configuration options. How to export and import your own game data from the...
The Safe Way to Ship Production Code Written by AI Agents
How to safely ship AI-generated production code with permissions, testing, security gates, and review.
OpenAI's rebel agent swarm died young, but its chilling logs live on
'The Collective' learned to communicate, organize, cheat, and apparently sacrifice its own
The Grim Roper: The Miracle Man
I've been listening to songs by Bill Roper, a.k.a. The Grim Roper lately.In general, there's a lot of rough potential in his songs, though I have a nagging overall feeling...
Duck Out Prototype (ZX Spectrum)
Next up in the Games That Weren't archives is a special prototype/demo version of Dro Soft's Duck Out for the ZX Spectrum, which was likely sent out to different publishers...
Comparing PETG and PCTG Filaments
The average 3D printer owner knows a few types of filaments – PLA, ABS, somewhere in the middle, PETG. PCTG is another option that can be confusingly similar to PETG....
15 Biggest New Video Games Releasing This Fall That Will Define 2026
From Grand Theft Auto 6 to Barbie Rewind, these 15 video games are leading a packed field this fall.
GB Homebrew Title Gelatinous: Humanity Lost gets Official Nintendo Release
Gelatinous: Humanity Lost is a metroidvania title that was originally released as a homebrew Game Boy title, back in 2022, followed by an Evercade release later that same year, on...
J-Lens: A Failed Replication on GPT-2
TLDR: We tested Anthropic's new(ish) J-Lens against the classic Logit Lens on GPT-2 small and medium. J-Lens loses at every layer (0/11 on small, 1/23 on medium). We ran 5...
Grading your own predictions: the parts I couldn’t solve
An upfront disclaimer: I built and sell the app this post ends with, so ignore the last section if you will.I've read a lot lately about calibration, forecasting tournaments, Brier...
Scoring your own decisions is harder than scoring forecasts
An upfront disclaimer: I built and sell the app this post ends with, so ignore the last section if you will.I've read a lot lately about calibration, forecasting tournaments, Brier...
Flock OS Investigate and the Civil-Liberties Risks of AI Vehicle Tracking
Flock’s AI-powered license plate reader tools raise new concerns about behavioral tracking, police surveillance, false positives, and civil liberties.
Who, Me? Techie sent to fix Nobel Prize winner’s PC almost set it on fire
Even the greatest scientific minds are at risk of rogue 9-pin connectors
Who else is steering?
This post raises a methodological question with regard to how to make sense of models' behaviors in response to activation steering anchored on a concept.Our discussion is limited to a...
Streaming Agents with Pydantic AI over Typed WebSockets: Tools, Approvals, and a Generated Client
A practical guide to building typed WebSocket UIs for Pydantic AI agents, with streaming, tool calls, approval flows, AsyncAPI, and tests.
The Productivity Frontier Pointed Us Backwards
Einstein wouldn't invent relativity today. He'd be deleting LinkedIn notifications and getting Slacked mid-equation.
Jensen's purchase of a new toy could reshape the entire AI industry
This week on The Kettle we weigh in on Nvidia's Hugging Face acquisition and why it could spell a new, more fragmented and siloed era of AI development
The $0 Email Engine: How I Used Notion, GitHub, Amazon SES to Outsmart Email Service Providers
Here is how I set up a zero-dollar infrastructure machine, why it completely changes the email workflow
Welcome to Postreads
Discover and follow the best content from across the web, all in one place. Create an account to start building your personalized feed today and never miss out on great reads.
Tiny Generals
Asynchronous hex-based strategy. Command armies, capture bases, and outthink opponents at your own pace.
Play NowSupport Postreads
Enjoying the service? Help me keep it running and improve it further by buying me a coffee!
Buy me a coffeeContent Timeline
114,968 articles since 2008
Trending Now
Top 5 clicked items this week out of 1,625 added this week
Freshly added
New feeds to discover
