Check out my #infosec #golang talk from @BSidesDetroit https://t.co/PxoJX03gWx

🦔 Static Taint Analysis for Go @KentGruber has been working on a static taint analysis package for @Golang → Can be a building block for detecting SQL injection, XSS, etc. In this post, he walks through how it works https://t.co/FFUTEHRnjK

Generative AI is a hot new topic, but how does it work? Find out on May 9th at 7PM where #misec Jackson will be hosting @KentGruber to present the primary challenges companies are facing building solutions on top of large language models.

So finally feel like I can talk about this. Santa as in https://t.co/1T0hNmP7lt now supports file access authorization. This means that we can authorize if a binary should be able to open a file/path and leverage code signing for targeting/filtering.

@gdb 😇

Plugins for processing a video clip, no ffmpeg wizardry required. Actual use-case from today's launch.

ChatGPT Plugins are here 🤯 Today we are launching: - Browsing in ChatGPT - Code execution in ChatGPT - 3rd party plugins And a whole new developer ecosystem! Here's what you need to know 🧵

Working on a way to model @OpenAI chat messages as a graph that can be summarized, searched, traversed, and more.

I found the JSONL format for the eval data to be a little cumbersome. Like, no VSCode syntax highlighting by default, for example. Each line can be very long, so they can be annoying to read or edit. So for fun, I decided to experiment with using HCL: https://t.co/oLaBXlKhkN

Currently, it has an accuracy of ~0.09 with GPT-3.5-Turbo. It can answer the example from the Wikipedia page, but struggles with the few others I generated. Super curious if GPT-4 will handle this better. https://t.co/r1mpxQd0lc

Spent some time learning about @OpenAI evals tonight, and submitted a Vigenère decryption evaluation. https://t.co/Xh0CkFzhfe

GPT-4 does drug discovery. Give it a currently available drug and it can: - Find compounds with similar properties - Modify them to make sure they're not patented - Purchase them from a supplier (even including sending an email with a purchase order)

it is more creative than previous models, it hallucinates significantly less, and it is less biased. it can pass a bar exam and score a 5 on several AP exams. there is a version with a 32k token context.

What it looks like without the ANSI bugs, the status bar at the bottom extends the full width. 🧵 Starts in a (filterable) list view to choose a “thread”.

📼 Used VHS to create this GIF. It didn’t come out perfect, but was really easy to use. I opened an issue to capture the problems I ran into. https://t.co/C0MaNik4tK

👩🏽‍💻 Continuing to figure out the UX for HAL. ✨ Enjoying learning about @charmcli’s Go packages.

🤩 Wow

HAL plugins to enable custom assistants, additional system configuration, commands, or any behavior. Allow plugins to be aware of the current context, summarizations provided by HAL and/or human. Plugins could also just change color of output, because that’s still fun.

Then there’s so many other goodies and tangential avenues to explore with this idea. Maybe I’ll sync my conversations, make them available across devices to save conversations with HAL. Could be compressed, segmented over time. Have HAL aware of program output, not just files.

There still so much to explore, but I want a multi-modal editor and shell controlled with natural language. Graph-based, preferably. With a bit of search and file management. It should help manage multiple lines of conversation/context, and limit tokens/messages. Cache history.