AI Chip Component Costs: Memory at 63% | Epoch AI
High-bandwidth memory (HBM) accounts for 63% of AI chip component costs, up from 52% in Q1 2024. Epoch AI's breakdown of component cost shifts across major chip designers.
davepeck.org
Dave Peck's home on the web. Dave is an independent software developer, investor, and civic technologist.
Garrett Peake
Software engineer building things on the internet.
“If it compiles, it works.” This feeling is one of the main things Rust engineers love most about Rust, and a reason why using it with coding agents is espec...
Watching for file changes on macOS
I've written a Swift script that uses the FSEvents API to get notified of file changes, then I'm using stdout as a bridge to forward those notifications to Python.
Why the gradient is a list of partial derivatives | Michał Prządka - Blog
Building the gradient formula from scratch using a ski-slope picture, with minimal calculus assumed.
Keep Your Slop to Yourself | David Gasquez
David Gasquez personal website
Extending MySQL with VillageSQL
One of the things that made me fall head over heals for Neo4j so many years ago was just how extensible it was. If the database engineering team was busy rebuilding the clustering feature for the t…
The Rise of the Backwards Deployed Engineer | Sagiv Ofek
The forward deployed engineer embeds with customers to drive revenue. The backwards deployed engineer embeds inward to eliminate SaaS bloat - and might be the most leveraged role in tech right now.
I have spent a lot of my career trying to envision and build better versions of the tech industry. To my face, people often say that I’m optimistic, though I suspect what they mean is that I’m engaged in an act of futility. I consider this work neither optimis…
Agents have good judgment now. Delegation beats micromanagement.
A reminder (to myself, mostly) on the importance of stepping away from the agents to make time for reflection
Claude doesn't know what time it is
Gemini has a clock. ChatGPT has the date. Claude has neither, and the obvious fix isn't as obvious as it looks.
agents don't actually want VMs
Every cloud provider is racing to sell agents a VM. But agents don't want a box to rent — they want a cloud to own. Here's why the next primitive is the Agent Cloud, and why Kubernetes has to be rebuilt to deliver it.
I am Not a Black Belt
I am a black belt. I am a second dan (nidan 二段) Aikido black belt with the original school of Aikido, the Aikikai foundation. I started Aikido practice in early 1997 after a holiday trip to the US. When Aikido was introduced in Brazil, it had inherited local r…
A personal reflection on programming after generative AI.
Buy from EU, but pay attention to what you buy.
At work, I have this growing anxiety that one of the LLM agents I'm encouraged to use is absolutely, positively going to delete something important and irretrievable.
One of the skillsets that the AI phenomenon has quietly boosted is the capability of explaining yourself: putting into words what you actually want, and providing methodologies that can unambiguously verify the implementation against your expectations. Turns o…
Aaron Swartz told us not to read the comments. He was right. The comment section of the early-to-mid internet was a place where nuance went to die and bad...
Recently, I have been running into the "This model is currently experiencing high demand" message using the Gemini API. This is frustrating since my side-pro...
Via John Markoff, I was invited to a conversation with Jonathan Dotan and the students of his EE292J course entitled Designing for Authent...
Before v0.128.0, every Codex CLI interaction ended the same way: the agent finished a turn, and you typed the next prompt. For a…
Reasonix — DeepSeek-native AI coding agent
Open-source AI coding agent for your terminal. Engineered around DeepSeek's prefix-cache. MCP first-class · plan mode · MIT.
Hello @filisavi (Member) ,
The C++ Standard Library Has Been Walking Itself Back for Fifteen Years, and the Receipts Are Public
# The C++ standard library has been walking itself back for fifteen years, and the receipts are public Sandor Dargo's [post this month on `std::copyable_function`](https://www.sandordargo.com/blog/2026/05/20/cpp26-copyable-function) closes with a quick-referen…
Amazon Web Services - Four Years and Out
Today marks four years since I joined AWS. My last day will be Friday. I have to say being fired from AWS is actually a relief. There have been a lot of changes to the company since I joined in 2022, and the company I wanted to work for is no longer the same c…
Declarative partial updates | Blog | Chrome for Developers
Learn about new out-of-order streaming capabilities and the renewed HTML insertion and streaming methods available for testing from Chrome 148
Released at the Outline Demoparty in May 2026, Ommen, NL An exploration of algorithmic density in 16 bytes of x86 assembly.
Italy moves to Airbus A330 tankers in major NATO-aligned shift
Rome shifts course: six Airbus A330 MRTT tanker aircraft, worth around €1.39 billion in total, to bolster the European pillar in NATO. #EuropeNews
Building a Minsky Machine in Atlassian Automation 22nd May 2026
Tips
It's time to talk about my writerdeck
tldr: * I installed console-only Debian on an old laptop (no desktop OS at all) * I added some packages to make it work better for writing: * network-manager for connecting to hotspots while I'm away * kmscon for custom fonts and more than 16 color…
Oura says it gets government demands for user data. Will it share how many?
Oura users' data is not end-to-end encrypted and can be handed to the government. Will the wearable tech maker say how often it turns over data?