Gemini Robotics 2 brings whole body intelligence to robots From feet to fingertips — we are teaching robots intelligent whole-body control, fine dexterity, and teamwork to complete a broad range of complex tasks.
Advancing the price-performance frontier with GPT-5.6 Explore lower GPT‑5.6 pricing for Luna and Terra—and how OpenAI’s more efficient models help enterprises deploy AI workflows at scale.
Performance Pitfalls in C# / .NET - Protobuf - repeated uint32 vs repeated fixed32 Looking at the performance of gRPC/protobuf serialisation
Everyone's Tables Suck Why is it so hard to add a cell? A tour of terrible tables in Notion, Google Docs, Linear, Obsidian, and Slack -- and how at Macro we're bringing tables back.
Why compute might get 10x more expensive in coming years If a human-level software engineer that could run on an H100 equivalent, at current market rates for software engineers, that H100 should rent for over $250k a year. That’s 15x today’s spot price.
Agency Is the Job Now Technical skill got repriced the day models could write most of the code. What's left is the willingness to blast through obstacles, and two stories about donuts and blood explain the whole thing.
We are surprisingly close to prompt-to-model in computer vision In computer vision, neural networks have become a commodity. You are a Google search away from downloading near-SOTA neural networks that are free, open source, have a permissive license, and are capable enough for most use cases.
We accidentally built an LLVM compiler for JAX | josh iza.ac Sometimes when you build software, you shave a yak so deeply that you end up with a surprisingly nice sweater (while the yak, presumably, wonders why it is suddenly rather drafty). For the past while, my team and I have been building a quantum compiler ca...
Why do OpenAI's GPT-2 weights beat mine? The original GPT-2 small weights are much better than my own models at a specific instruction-following task. Time to look into why!
I Tested Six Ways to Save $$$ on 5.6 Sol. The only one that worked was using Terra Six interventions meant to save tokens gave different answers depending on how you count. None cut the ten-task batch total twice. Caveman and the cheaper Terra model looked mildly cheaper per task, but within run-to-run noise.
I want to leave the internet I want to leave the internet. This site, reddit, my banking app, the cloud service, Youtube, my music streaming app, my email, Whatsapp, iNaturalist, pornsit...
headroom cut 39% of my tokens and raised my claude bill headroom cut 39% of input tokens off an agentic coding benchmark and the bill, if anything, went up. prompt caching is why, and most savings meters cannot see it.
When Reading Roman History I finished reading SPQR by Mary Beard a few months back. In this overdue post, I share things that are important to keep in mind when casual history nerds like myself read about Roman history and ancient history more generally.
Why Performance Improvements Compound Mature systems rarely waste time inside algorithms. They waste time moving data between them—and that is why performance fixes compound.
That time when I failed the Microsoft interview The year was 2015, halfway through my Computer Science bachelor’s degree. I had spent a few months contributing to open source and my confidence levels were high, so I thought: why not aim for a summer internship at Microsoft? Microsoft, of all places, mi...
Why are databases so hard? You've probably all experienced it; another outage and the database is the root cause. Why are databases such a frequent cause of problems ...
What's That Burning Smell? - Off Grid and Free: My Path to the Wilderness I share a lesson learned for anybody who has an off grid solar electric system. A problem - a solution and the math behind it all.
Software for One - Adam Waxman Robin Sloan wished for a HyperCard that could build a family app in a day. That world showed up.
LLMs and shaders In the last post I described how chat interfaces to LLMs are giving a limited view of what the LLMs could do. In this post I wanted to com...
Distributed rate limiter with HRW in Elixir | jola.dev Building the classic distributed rate limiter, using the HRW library instead of the traditional ExHashRing.
How Real-Time Apps Actually Work We use real-time apps every day but rarely stop to think about how they work.
