What we're reading, in the order we're reading it.
Anthony and Harry's working stream — links, charts, tweets, and short takes. The unprocessed inputs behind the briefings.
Predicting water quality index using stacked ensemble regression and SHAP based explainable artificial intelligence
Scientific Reports – Predicting water quality index using stacked ensemble regression and SHAP based explainable artificial intelligence
The Hidden Ingredients Behind AI’s Creativity
Image generators are designed to mimic their training data, so where does their apparent creativity come from? A recent study suggests that it’s an inevitable by-product of their architecture.
How to Fix Your Context
6 tactics for fixing your context and shipping better agents. As Karpathy says, building LLM-powered apps means learning to ‘pack the context windows just right’—smartly deploying tools, managing information, and...
How artificial intelligence uncouples hard work from fair wages through ‘surveillance pay’ practices—and how to fix it
How surveillance pay practices work, where they are increasingly deployed in the U.S. economy, and policy recommendations to ensure pay fairness and transparency.
GitHub – martianlantern/ThinkMesh: Parallel thinking for LLMs. Confidence‑gated, strategy‑driven, offline‑friendly
Parallel thinking for LLMs. Confidence‑gated, strategy‑driven, offline‑friendly – martianlantern/ThinkMesh
Agentic Browser Security: Indirect Prompt Injection in Perplexity Comet
The attack we developed shows that traditional Web security assumptions don’t hold for agentic AI, and that we need new security and privacy architectures for agentic browsing.
AGI is an Engineering Problem
LLM models are plateauing, but true AGI isn’t about scaling the next breakthrough model—it’s about engineering the right context, memory, and workflow systems. AGI is fundamentally a systems engineering problem,...
Unitree G1 – ROBOTS: Your Guide to the World of Robotics
The Unitree G1 is a compact humanoid robot designed for AI-powered mobility and manipulation. It stands out for its agility, force-controlled dexterous hands, and the ability to perform tasks like...
Principal Software Engineer
About Motion Given the speed of AI advancement, we operate under two assumptions about the future of work: 1. AI will do 100x more work over the next decade 2....
AI-Powered Robot by Boston Dynamics and Toyota Research Institute Takes a Key Step Towards General-Purpose Humanoids
Joint research collaboration enables the Atlas humanoid robot to achieve autonomous whole-body manipulation and locomotion behaviors using Large Behavior Models
Politicians Are Trying to Make It Illegal to Sue AI Companies
A new bill in Colorado would prevent individuals from suing AI companies under the state’s consumer protection laws.
Building a16z’s Personal AI Workstation with four NVIDIA RTX 6000 Pro Blackwell Max-Q GPUs
In the era of foundation models, multimodal AI, LLMs, and ever-larger datasets, access to raw compute is still one of the biggest bottlenecks for researchers, founders, developers, and engineers. While...
6 Things To Know About The Meta/Google Cloud AI Deal
Meta struck a massive $10B AI cloud hosting deal with Google, boosting Google’s cloud revenue and solidifying Meta’s AI expansion plans.
Tests Show That Top AI Models Are Making Disastrous Errors When Used for Journalism
Journalists found in an investigation that AI is strikingly terrible at summarizing documents and scientific research for busy reporters.
GitHub – ayushmi/agentstate: Cloud-native, durable state for AI agents: WAL+snapshots, watch streams, idempotency, leases, TLS/mTLS, capability tokens, Python/TS SDKs, Helm.
Cloud-native, durable state for AI agents: WAL+snapshots, watch streams, idempotency, leases, TLS/mTLS, capability tokens, Python/TS SDKs, Helm. – ayushmi/agentstate
On the use of LLM assistants for kernel development
By some appearances, at least, the kernel community has been relatively insulated from the onsl […]
This Week in AI: AI Bubble Fears Tank Tech Stocks
Some of the biggest names in tech, including Google, Amazon, and Apple, saw their stocks battered this week amid fears that the artificial intelligence
OpenCUA’s open source computer-use agents rival proprietary models from OpenAI and Anthropic
The open source framework provides the data and training recipe for building powerful computer-use agents that challenge proprietary systems.
Meta is partnering with Midjourney and will license its technology for ‘future models and products’
Details are scarce on how much if any money is involved. And what does it mean for Midjourney’s previously announced plans for an enterprise API?
Apple Explores Using Google Gemini AI to Power Revamped Siri
Apple Inc. is in early discussions about using Google Gemini to power a revamped version of the Siri voice assistant, marking a key potential step toward outsourcing more of its...
MCP-Universe benchmark shows GPT-5 fails more than half of real-world orchestration tasks
A new benchmark from Salesforce research evaluates model and agentic performance on real-life enterprise tasks.
Artificial intelligence bill clears first Senate special session committee
The proposed Senate Democrats’ fix to Senate Bill 24-205, the first-in-the-nation artificial intelligence regulation on discrimination, won approval from its first committee on the first day of the 2025 special
Essential Reading for Agentic Engineers – August 2025
Five essential perspectives that cut through AI hype: from developer evolution stages to junior learning crises, productivity reality checks, platform disruption, and MCP server pitfalls
“Velocity IS The ****ing Strategy”: What Citizen Development Means For AI-Enhanced Businesses
AI continue to be hard. Our research (and others) shows that even deploying a single, substantive genAI solution or agent is exceptional. There are two anti-patterns that complicate adoption: first,...
Harnessing artificial intelligence to identify Bufalin as a molecular glue degrader of estrogen receptor alpha
Nature Communications – Target identification of natural products plays a critical role in the development of innovative drugs. Here, this group integrates molecular docking and molecular dynamics…
The new Hollywood: Runway’s AI models edge closer to simulating reality
Its models were among the first to generate synthetic video for movies like “Everything Everywhere All at Once.” Now they can mirror the laws of physics.
Why Nvidia Earnings Matter More To Markets Than What The Fed Chair Said
We know where the Fed chair stands. But uncertainty over Nvidia’s earnings could worry investors concerned about an AI bubble.
Apple in talks to use Google’s Gemini AI to power revamped Siri, Bloomberg News reports
Apple is in early talks to use Google’s Gemini AI to revamp the Siri voice assistant, Bloomberg News reported on Friday, citing people familiar with the matter.
Apple considers Google Gemini to power next-gen Siri, internal AI ‘bake-off’ underway
Apple seems open to anything and everything when it comes to delivering the next generation of Siri. After reports that…
AI Update, August 22, 2025: AI News and Views From the Past Week
Artificial Intelligence – Catch up on select AI news and developments from the past week or so. Stay in the know.
How to orchestrate AI workflows in 7 steps
In this guide, you’ll get a repeatable, 7-step framework that helps your teams orchestrate smarter systems using tools like Zapier Canvas, Zapier Tables, and Zapier Agents.
95% of business applications of AI have failed. Here’s why
Here’s what the other 5% is getting right.
Google has an all-new Gemini AI service built specially for the US Government
Google reveals ‘Gemini for Government’
AI fails to bring businesses returns, report says
The MIT research found that, of more than 300 organizations analyzed, 95% saw zero financial returns for their AI efforts.
Being “Confidently Wrong” is holding AI back
The failure mode that stalls “AI for data” efforts or “AI on my APIs” efforts isn’t psychedelic hallucination—it’s confident inaccuracy: plausible answers that are wrong in subtle and costly ways.
Meet the researcher hosting a scientific conference by and for AI
James Zou used virtual AI ‘scientists’ to find novel covid-19 treatments. Now, he’s having AI review and present all of the research at a controversial new conference.
The AI Model Most Favored by Enterprises Is Not What You Think
OpenAI’s ChatGPT took the world by the proverbial storm when it debuted in 2022, becoming synonymous with artificial intelligence (AI) in the public eye.
South Korea makes AI investment a top policy priority to support flagging growth
South Korea vowed on Friday to make investment in artificial intelligence a top policy priority, as the government slashed its economic growth projection for this year due to trade headwinds...
UMD survey finds bipartisan support for AI regulation
Recent surveys have found people don’t want innovation stifled by AI regulation, but are more worried about unchecked development of AI systems.
Don’t sleep on Cohere: Command A Reasoning, its first reasoning model, is built for enterprise customer service and more
It looks to be a strong release. Benchmarks, technical specs, and early tests suggest the model delivers on flexibility, efficiency, and raw