I designed my first real part!
www.printables.com/model/162187...
No, I don't really know what I'm doing. Yes, it is satisfying anyway!
01.03.2026 15:26
👍 0
🔁 0
💬 0
📌 0
I designed my first real part!
www.printables.com/model/162187...
No, I don't really know what I'm doing. Yes, it is satisfying anyway!
Ooh that sounds like a very interesting talk. Verification and test-driven development are going to be very important in the near future.
TIL about pypi-timemachine. You're welcome. (You know, for those 10-year old research projects without lock files)
snap: When you only care about 90% of your apps to work.
It's been years. Why do major snaps (firefox!) still have usability issues?!
openai.com/index/harnes...
"What’s become clear: building software still demands discipline, but the discipline shows up more in the scaffolding rather than the code. The tooling, abstractions, and feedback loops that keep the codebase coherent are increasingly important."
x.com/mrexodia/sta...
I concur with most of this advice on vibe coding.
I've seen a few of these AI skim tools, but I like Google scholar's new one: scholar.googleblog.com/2024/11/ai-o...
TIL that fine-tuning a PEFT adapter for a pretrained model (i.e., not a model fine-tuned for chat) is probably a bad idea. By default, PEFT adapters don't include the vocab embeddings, so it is more or less unable to learn the meaning of tokens not in pretraining like EOS. Whoops!
🚨 Blog Post: ""Idioms: A Simple and Effective Framework for Turbo-Charging Local Neural Decompilation with Well-Define... https://edmcman.github.io/blog/2026-01-15--idioms-a-simple-and-effective-framework-for-turbo-charging-local-neural-decompilation-with-well-defined-types-accepted-to-ndss-2026/
Reflecting on the success of our first SURE and beginning the planning for the next year!
Zscaler of course was another problem, EVEN WHEN IT WAS TURNED OFF! Yesterday's image is confirmed.
Some days I hate computers. Today I've been debugging a packer build of windows 11 arm using vmware fusion, which is uncharted territory already. There were many problems. But the last and most frustrating was Mac silently blocking access to my "local network" bc I was in VS Code. 🤯
Today's cool visualization of the day is brought to you by arxiv.org/pdf/2512.14045
The world needs more Sankey diagrams.
I always felt like tqdm's ETA estimations were wildly inaccurate. It's because it defaults to using an exponentially weighted moving average with 0.3 weight. With high variance in job times and a lot of threads, that isn't going to work well.
tqdm.github.io/docs/tqdm/#:....
Not a good vibe coding day. GH Co-pilot couldn't seem to figure out how to read a file.
Problem 1: Task was given to sub-agent without the path. Oops.
Problem 2: I had so many files in the workspace that search was timing out.
I think HF uses AWS ECS under the hood. Not sure if they have any other type of isolation.
No :-( Under the hood, HF spaces are just Docker containers, so a permanent image could be distributed that way.
varlogsimon.leaflet.pub/3m6zrw6k2bs2p
Back in my day, TLS was considered end to end encryption.
Who did you think was going to be on the other "end" of your toilet besides Kohler? Your social contacts? 🚽
🚨 Blog Post: "Musings on Decompilation Artifacts" https://edmcman.github.io/blog/2025-12-06--musings-on-decompilation-artifacts/
Must-read for fuzzing folks (read: tooling/algorithms/academia) by Addison Crump
addisoncrump.info/research/wha...
You've seen the trends in AIxCC: LLMs can hack source, find vulns, and patch them. But what about on binaries without source? Do decompilers close the gap, or is there more to grow?
Come see my talk at DistrctCon where I merge and dissect these two fields: AI Hacking + Decomp.
