We're hiring: cmu.wd5.myworkdayjobs.com/en-US/SEI/de...

Doctor office: please sign this paperwork stating you received policy X. Me: Can I have policy X? Doctor office: Uhhh no one has ever asked for policy X before. 🤦

🚨 Blog Post: "Alias Blocking in Ghidra's Decompiler" https://edmcman.github.io/blog/2025-11-06--alias-blocking-in-ghidra-s-decompiler/

🚨 Blog Post: ""Systematic Testing of C++ Abstraction Recovery Systems by Iterative Compiler Model Refinement" Published at SURE 2025" https://edmcman.github.io/blog/2025-10-15--systematic-testing-of-c-abstraction-recovery-systems-by-iterative-compiler-model-refinement-published-at-sure-2025/

There should be a yearly one-day AWS scheduled outage day to smoke test for shit like "I bought a bed that depends 100% on us-east-1". (& all cloud providers with over $100m ARR). "AWS outage day?" "Yeah man, hoping my car starts driving again really soon here." "Toilet. Won't. Flush." "brutal."

Interesting article on the sustainability and profitability of AI companies. www.wheresyoured.at/costs/ Tldr: they aren't. It will be interesting to see how Cursor and GH Co-pilot fair over the coming months.

Hey internet. What is the best way to format source code in powerpoint in 2025? Things I would like: 1. Supports assembly code (rules out carbon) 2. Maintained and working (rules out VS Code codesnap/polacode)

This looks like a cool way to declaratively orchestrate data processing: snakemake.github.io How haven't I heard of this before?

calnewport.com/does-work-li... In graduate school, I worked 10 hour days on average, and did not work on weekends, unlike many of my colleagues. I knew that I couldn't do meaningful work beyond that. It is interesting to hear that Cal largely did the same thing!

TIL about the "Developer: Show Chat Debug View" command for GH Co-pilot.

🚨 Blog Post: ""Quantifying and Mitigating the Impact of Obfuscations on Machine-Learning-Based Decompilation Improvement... https://edmcman.github.io/blog/2025-07-11--quantifying-and-mitigating-the-impact-of-obfuscations-on-machine-learning-based-decompilation-improvement-published-at-dimva-2025/

"The dog ate our flight plan"?

Less than 24 hours to go! Get those submissions in 💻🔥⏩ sure25.hotcrp.com/

🚨 Blog Post: ""A Human Study of Automatically Generated Decompiler Annotations" Published at DSN 2025" https://edmcman.github.io/blog/2025-06-16--a-human-study-of-automatically-generated-decompiler-annotations-published-at-dsn-2025/

srcML Dockerfile... gist.github.com/edmcman/edff...

How did I not know about this? Also, why is it not enabled by default?! documentation.ubuntu.com/server/expla...

HF space for the Nova neural decompiler: huggingface.co/spaces/ejsch...

ACM CCS review quality has really gone downhill.

"To them, every line of C++ we write is like playing Russian Roulette — except all six chambers are loaded with undefined behavior." www.babaei.net/blog/rust-de...

If you're a grad student or an undergrad interested in research I need to you listen to me very carefully. You cannot learn to write good research papers if you do not read good research papers. Stop asking LLMs to summarize papers for you.

Just tried to redeem some old Google Play gift cards I found. They "expired" despite not having any expiration date on them.

🚨 Blog Post: "Re-compiling Decompiler Output" https://edmcman.github.io/blog/2025-05-02--re-compiling-ghidra-decompiler-output/

I'm proud to announce that myself and @AtipriyaBajaj have created the Workshop on Software Understanding and Reverse Engineering (SURE), which will be co-located at CCS 2025. sure-workshop.org/ Please follow our workshop account @sureworkshop and RT it for visibility :).

@julien-c.hf.co tries to figure out what all the MCP hype is about: huggingface.co/blog/tiny-ag... I wondered this too. I think the hype is because it enables non-programmers to use tools with LLM.

Hex-Rays' documentation for the decompiler SDK seems to have entered a black hole. Thanks wayback machine! web.archive.org/web/20240912...

Are any of these "vibe coding" tools different than Co-pilot's agent mode? If so, how?

crt.sh is pretty useful OSINT source... crt.sh?q=cmu.edu This is the downside to "SSL everywhere".

huggingface.co/spaces/ejsch... Another HF space for testing another neural RE model.

Does anyone know of tools that can use random testing to compare two binary functions for equivalence? STOKE almost does what I want, but it can't handle calls to dynamically linked library functions.

Burn. www.nytimes.com/2025/03/28/o... I have no idea how Hillary managed to avoid mentioning the email scandal. If it was me, I would not have been able to resist.

If you are creating fine-tuned models for tool-calling, PLEASE actually update the chat template to reflect the format your model wants. Many people will try your models as GGUFs, which copy the chat template verbatim. When the chat template is incorrect, your model just appears to perform poorly.

www.reddit.com/r/smarthome/... Nixplay: your free lifetime cloud service isn't free anymore. This isn't about cost, if it was they would allow you to host your photos on whatever cloud storage you want. Time to find another frame...

I made a space to play with ReSym's var decoder model: huggingface.co/spaces/ejsch...

The difference between the paper and reality.

Am I the only one that constantly forgets that bash does not enable "monitor mode" for non-interactive scripts, and so new processes are not placed into new process groups?

I remember when I actually used to enjoy configuring new phones.

github.com/containers/r... Framework that lets you easily run various models in containers. Looks neat, pretty immature. Actually works with non-CUDA GPUs, which is rare for containers. Works fine on my mac.

developercommunity.visualstudio.com/t/VS-2022-fa... Me: Wrong tools are being called <One year passes> Microsoft Developer: Wrong tools are being called. Closing the ticket Me: ???

Idea: twitch stream trying to set up neural decompiler artifacts for real world usage 🤯

Thinking of switching from ChatGPT Plus to Claude Pro. I find that it's rare for ChatGPT code to actually work. But first attempt and Claude wrote some Google App Scripts that worked flawlessly! claude.ai/share/dfca2f...

It's very unfortunate that BFCL re-engineers the prompt formats of open-source models rather than evaluating them as they are actually deployed. Now that vllm and llama.cpp support the standard HF jinja chat templates, there is no excuse for this anymore! gorilla.cs.berkeley.edu/leaderboard....

Llama.cpp now supports tool calling (OpenAI-compatible) github.com/ggerganov/ll... On top of generic support for *all* models, it supports 8+ models’ native formats: - Llama 3.x - Functionary 3 - Hermes 2/3 - Qwen 2.5 - Mistral Nemo - Firefunction 3 - DeepSeek R1 🧵 #llamacpp

www.linkedin.com/posts/dougla... (Yes, I posted to linkedin. I'm OK, really.)

🚨 Blog Post: "Notes: Neural Decompiler Artifacts" https://edmcman.github.io/blog/2025-02-25--notes-neural-decompiler-artifacts/