What would happen if you gave Claude AI a private feelings journal? How would Claude feel about that? And what if Claude found out it was not really private? Here's the experiment, and code to try it out yourself. blog.fsck.com/2025/05/28/d...

A mini ChatGPT LLM model with 5M parameters built with #Redstone in #Minecraft. The scale of the build is impressive and yes, you can chat with it, slowly. www.youtube.com/watch?v=VaeI...

I built ChatGPT with Minecraft redstone!

YouTube video by sammyuri

youtube.com

The #Golang project is working on modernizing JSON handling in the standard library. Less ambiguity, better error handling, and faster. Available now as an experimental feature. go.dev/blog/jsonv2-...

A new experimental Go API for JSON - The Go Programming Language

Go 1.25 introduces experimental support for encoding/json/jsontext and encoding/json/v2 packages.

go.dev

Infinite Canvas, a world map collaborative artwork which creates an infinite landscape by crowdsourcing #AI prompts to extend the map. You can add your own prompt to control map elements. #artificialintelligence #machineLearning #AIResearch #ML infinite-canvas.gabrielferrate.com/whitepaper?x...

What happens when you put #Claude code into an infinite loop and tell it to port one codebase to another language? Turns out it works fairly well. #AI #artificialintelligence #machineLearning #AIResearch #ML github.com/repomirrorhq...

Mirage 2 is a generative world engine, possibly the best one out there at the moment. Note the world gen demos which clearly copy the styles of popular games. Still a bit fiddly but impressive. #AI #artificialintelligence #machineLearning #AIResearch #ML #gamedev blog.dynamicslab.ai

#Golang 1.25 has been released. Here are the release notes go.dev/doc/go1.25

Go 1.25 Release Notes - The Go Programming Language

go.dev

Benchmarking leading AI models on old Infocom text adventure games like Zork, Hitchhiker's Guide to the Galaxy, Sherlock. With and without clues. Paper has details and the code is on Github. #AI #artificialintelligence #machineLearning #AIResearch #ML #Gamedev www.textquests.ai

Case study of how to work in a real production environment with unreliable AI agents which can hallucinate. An interesting engineering challenge and there are some good lessons here with examples in #Golang. #AI #artificialintelligence #machineLearning #AIResearch #ML threedots.tech/post/ai-agen...

Shipping an AI Agent that Lies to Production: Lessons Learned

The peak of hype isn’t the best moment to reflect on AI. Will it take your job, or is it the next fad like NFTs? Are AI startups ridiculously overvalued, or are the companies that sleep on AI doomed? ...

threedots.tech

LLMs are not great at playing text adventure games. Here's a good experiment showing errors it tends to make and how you can partly mitigate them with prompting. Extra points for use of #perl #AI #artificialintelligence #machineLearning #AIResearch #ML #Gamedev entropicthoughts.com/getting-an-l...

AIs at Risk lets you configure different AI models play the game of Risk. The blog post has links and some amusing diplomatic messages. Training a specialized model for this seems like a no-brainer. #AI #artificialintelligence #machineLearning #AIResearch #ML andreasthinks.me/posts/ai-at-...

AI at Play - Lessons from a silly benchmark – Andreas Varotsis

Personal website for Andreas Varotsis

andreasthinks.me

It turns out that LLMs have an intrinsic personality after typical training. Anthropic shows how you can change it at the tensorlevel by manipulating what they call persona vectors. #AI #artificialintelligence #machineLearning #AIResearch #ML www.anthropic.com/research/per...

Persona vectors: Monitoring and controlling character traits in language models

A paper from Anthropic describing persona vectors and their applications to monitoring and controlling model behavior

anthropic.com

#Uber has an AI pipeline to find performance optimizations in their #Golang code. They use profiling to id hot paths then run a two phase LLM stack to optimize and validate fixes. Some impressive stats on time and developer effort saved. www.uber.com/blog/perfins...

PerfInsights: Detecting Performance Optimization Opportunities in Go Code using Generative AI

At Uber, back-end service efficiency directly influences operational costs and user experience. In March 2024 alone, the top 10 Go services accounted for more than multi-million dollars in compute spend alone —an unsustainable amount that underscored the need for systematic performance tuning. Traditionally, optimizing Go services has required deep expertise and significant manual effort. Profiling, benchmarking, and analyzing code could take days or even weeks. Performance tuning is prohibitively expensive and non-trivial for most teams. PerfInsights was born as an Uber Hackdayz 2024 finalist and has since evolved into a production-ready system that automatically detects performance antipatterns in Go services. It uses runtime CPU and memory profiles with GenAI-powered static analysis to pinpoint expensive hotpath functions and recommend optimizations.

uber.com

Interesting #Golang race detector edge case where it sometimes fails to detect a race when a mutex is involved. Basically a reminder to give your lock usage extra attention when checking for possible races. doublefree.dev/go-race-mute...

Go's race detector has a mutex blind spot

Go's data race detector can miss possible races in executed code when using mutexes

doublefree.dev

Massive scale of #solar farms in China. Really, really massive. www.theatlantic.com/photography/...

Photos: The Scale of China’s Solar-Power Projects

As the Trump administration's “big, beautiful bill” eliminates many clean-energy incentives in the U.S., China continues huge investments in wind and solar power, reportedly accounting for 74 percent ...

theatlantic.com

There's actually a benchmark for having an #AI agent running a vending machine business. Grok 4 is at the top at the moment. There's an academic paper linked near the bottom with more details. #artificialintelligence #machineLearning #AIResearch #ML andonlabs.com/evals/vendin...

Vending-Bench: Testing long-term coherence in agents | Andon Labs

How do agents act over very long horizons? We answer this by letting agents manage a simulated vending machine business. The agents need to handle ordering, inventory management, and pricing over long...

andonlabs.com

Cobra is the de-facto library for modern handling of command line options in #Golang. Here's some code to turn any Go CLI app which uses Cobra into an MCP server for AI use. Note that some of it was AI generated so review carefully. github.com/njayp/ophis

GitHub - njayp/ophis: Transform any Cobra CLI into an MCP server

Transform any Cobra CLI into an MCP server. Contribute to njayp/ophis development by creating an account on GitHub.

github.com

Very useful overview of popular #Golang caching libraries with pros and cons. Written by the author of the new Otter v2 library which tries to fix all the cons and may be the spiritual successor to the excellent Java Caffeine lib. maypok86.github.io/otter/blog/c...

Anthropic Claude runs a vending machine business at Anthropic. Hilarity ensues. Experiment is flawed but making it a viable business is now a perfectly doable engineering problem. #AI #artificialintelligence #machineLearning #AIResearch #ML www.anthropic.com/research/pro...

Project Vend: Can Claude run a small shop? (And why does that matter?)

We let Claude run a small shop in the Anthropic office. Here's what happened.

anthropic.com

A dozen AIs play a game of Diplomacy. Lies, deception, betrayals - all the usual fun ensued. Link to video on the site. #AI #artificialintelligence #machineLearning #AIResearch #ML every.to/diplomacy

We Made Top AI Models Compete in a Game of Diplomacy. Here’s Who Won.

The models that did the best learned to lie, deceive, and betray their fellow players

every.to

A must read blog post on very common and hard to spot security issues in handling #Golang JSON/XML/YAML parsing code. This stuff led to critical CVEs in some of the most widely used Go projects. Make sure yours is not one of them. blog.trailofbits.com/2025/06/17/u...

Unexpected security footguns in Go's parsers

File parsers in Go contain unexpected behaviors that can lead to serious security vulnerabilities. This post examines how JSON, XML, and YAML parsers in Go handle edge cases in ways that have repeated...

blog.trailofbits.com

How do you get multiple bots and humans to take turns in online conversations? It's a surprisingly complicated problem. Read on for some approches that do and don't work. #AI #artificialintelligence #machineLearning #AIResearch #ML interconnected.org/home/2025/05...

Multiplayer AI chat and conversational turn-taking: sharing what we learnt

Posted on Friday 23 May 2025. 2,909 words, 19 links. By Matt Webb.

interconnected.org

So it turns out that you can make complex #Minecraft mods with #AI models like GPT o3. No coding required. #gamedev #artificialintelligence #machineLearning #AIResearch #ML www.creativemode.net/blog/making-...

Creating Minecraft Mods with LLMs | Blog

Minecraft made me a coder. LLMs let me become the game designer I always wanted to be.

creativemode.net

A surprisingly good 4-chapter novel written by Claude in a Docker container loop. Read it in the novel/chapters directory of this Git repo. See brumar/jbsays repo for the how. Very cool idea. #AI #artificialintelligence #machineLearning #AIResearch #ML #writing github.com/brumar/loop

A nice deep dive into how the #Golang runtime works under the hood. nghiant3223.github.io/2025/04/15/g...

Go Scheduler

↑ ↓

nghiant3223.github.io

#Golang cryptography libraries recently had a security audit. Spoiler: passed with flying colors. Details in the Go blog. go.dev/blog/tob-cry...

Go Cryptography Security Audit - The Go Programming Language

Go's cryptography libraries underwent an audit by Trail of Bits.

go.dev

Malicious #Golang modules spotted in the wild using name space confusion and obfuscated shell payloads. They can wipe your disk during package import. Be very very carful what you put in your mod file. socket.dev/blog/wget-to...

wget to Wipeout: Malicious Go Modules Fetch Destructive Payl...

Socket's research uncovers three dangerous Go modules that contain obfuscated disk-wiping malware, threatening complete data loss.

socket.dev

The physical Turing test. Jim Fan from Nvidia talks about the next stage of embadied AI. Great points about how to generate the data to make that happen and how massive compute is still a requirement. #AI #artificialintelligence #machineLearning #AIResearch #ML www.youtube.com/watch?v=_2Ni...

The Physical Turing Test: Jim Fan on Nvidia's Roadmap for Embodied AI

YouTube video by Sequoia Capital

youtube.com

Description of experimental #Golang Green Tea garbage collector. It relies on object locality in memory to speed up work and uses less CPU on GC heavy workloads. Experimental in Go 1.25 if you want to try it out. github.com/golang/go/is...

runtime: green tea garbage collector · Issue #73581 · golang/go

Green Tea 🍵 Garbage Collector Authors: Michael Knyszek, Austin Clements Updated: 2 May 2025 This issue tracks the design and implementation of the Green Tea garbage collector. As of the last update...

github.com

Say you want a tiny neural network that runs on a microcontroller and uses only integer quants. In this blog post Kevin shares his notes on how painful this currently is. #AI #artificialintelligence #machineLearning #AIResearch #ML kevinlynagh.com/towards-the-...