Great summary! I would just add that cherry on top though -- that HuggingFace tried using the top commercial models in response but couldn't because of the cybersecurity restrictions so they had to use GLM 5.2 instead
"When we started the log analysis, we first used frontier models behind commercial APIs. This did not work: the analysis requires submitting large volumes of real attack commands, exploit payloads, and C2 artifacts, and these requests were blocked by the providers' safety guardrails, which cannot distinguish an incident responder from an attacker. We ran the forensic analysis instead on GLM 5.2, an open-weight model, on our own infrastructure. This had a second benefit: no attacker data, and none of the credentials it referenced, left our environment."
Heh after reading that title card I thought this was going to be a mosquito based software bug analogy. I expected a description of how to write software that resulted in more "good bugs" that might facilitate finding other bugs somehow. Now I'm a little disappointed
It’s just a basic IntelliJ plugin which provides an infinite canvas to add code bookmarks to. I work on a large code base and often have to take on tasks involving lots of unfamiliar areas of code and components which influence each other only through long chains of indirection. Having a visual space to lay things out, draw connections, and quickly jump back into the code has been really helpful
The canvas and UI is built using Java AWT since that’s what IntelliJ plugins are built on, but it occurred to me that I could just throw in a web view and use any of the existing JS libraries for working on an infinite canvas. React Flow has seemed like the best option with tldraw being what I’d fallback to.
But then.. if the canvas is built with web technology then there’s no reason to keep it just within an IntelliJ plugin vs just a standalone web app with the ability to contain generic content that might open files in IntelliJ or any other editor. I’m pretty sure the “knowledge database on a canvas” thing has been done a number of times already so I want to also see if there are existing open source projects that it’d be easy enough to just add a special node type to
See also this recent post about Mercury-Coder from Inception Labs. There's a "diffusion effect" toggle for their chat interface but I have no idea if that's an accurate representation of the model's diffusion process or just some randomly generated characters showing what the diffusion process looks like
Just about every line of code for this was written by Claude 3.7 via Claude Code. I never gave the other AI development tools like Cursor and Aider a fair shake so I may have just been behind the times on what's possible with agentic editors, but I found Claude Code to be extremely impressive. The API costs did end up being something around $15 though and this is a really small project, so I imagine cost goes up quite a bit to do any non-trivial work on a project of a decent size
This was great for me though because this is something I've had in mind for a while but never considered quite useful enough to be worth the time it'd take to learn about IntelliJ plugin development + the Java AWT knowledge required to create a canvas app like this
lol yeah I tried to get it to whisper too. And talk faster or slower or do accents. It seemed to be able to kind of do each of those things but only very slightly. Enough to see that there was some successful interpretation of the request but lack of flexibility to fully execute on it. OpenAI's model still has this beat on that front imo (talking quietly / slower / faster)
Yeah after a few interactions, the repetition of the mannerisms that initially added to the sense of life-likeness started to break the illusion a bit. The "you got me" response shows up a bit too often. The creativity remains impressive though
This was already posted here: https://news.ycombinator.com/item?id=43221377 but I’m really surprised at the lack of attention this model is getting. The responsiveness and apparent personality are pretty mind blowing. It’s similar to what OpenAI had initially demoed for advanced voice mode, at least for the voice conversation portion.
The demo interactions are recorded, which is mentioned in their disclaimer under the demo UI. What isn't mentioned though is that they include past conversations in the context for the model on future interactions. It was pretty surprising to be greeted with something like "welcome back" and the model being able to reference what was said in previous interactions. The full disclaimer on the page for the demo is:
"
1. Microphone permission is required. 2. Calls are recorded for quality review but not used for ML training and are deleted within 30 days. 3. By using this demo, you are agreeing to our
"
A cross-platform clipboard manager / search-and-filter tool / launcher built with Flutter that has a simple Python plugin interface.
Plugins can be used to add new "result actions" and new sources of entries to filter and select. Eg. recent Jira tickets, email inbox, shell history, Notion pages, etc.
The result actions are a way to easily perform common transformations on selected entries (eg. wrap in triple backticks, find and parse json, trim whitespace, ...) or kickoff some script with a selected entry as an argument.
Project started as a result of having to do a lot of work using Ubuntu and sorely missing Alfred and all the workflows I'd built with it. I wanted something for which I could build workflows once and have those workflows available on whatever system I'm on. Plus to be able to build some plugins that would be usable by coworkers regardless of what operating system they're using and with minimal runtime resource usage. There are some existing cross-platform solutions which could serve this purpose, like Cerebro, Ueli, Script Kit, some others.., but I wanted something lighter weight than is possible with an Electron app. Granted the current state of Epte is that it's built with Flutter + Go + Python so the final distributable and runtime memory usage are higher than is ideal.
Basic Windows support is almost there but there doesn't seem to be a great solution to switching to existing windows of an application instead of just re-launching it. The tool isn't intended to be as good or better than any given OS's built-in launcher so I'll probably just leave that as-is and upload the current state of the Windows build.
The best approach I’ve found so far is to just have a single master “event log” where I dump everything that I want to save by default. I have specific places to put things but if I can’t be bothered to decide where or am not sure it’ll just go to the event log. I’m using Notion for this where each entry is its own page in a “database” list. Adding a new page is trivial though through the site or app. I have an iOS shortcut setup too to open the entry creation
Anyone else have their zsh history get truncated occasionally? Not sure where in my setup the issue is but sometimes the history file will just be recreated. I have all the histsize and other env vars set correctly and I haven’t been able to correlate it to anything in particular I’m doing when it happens
If servo (or something like it) succeeds, would that mean potentially being able to swap out chromium in Electron? Would that help with performance / application size?
I've been working on a cross-platform clipboard manager launcher thing and using xdotool to simulate a ctrl+v on Ubuntu. On every paste though, the entire screen would freeze for a solid second. After a lot of exploring options I came across https://github.com/robertwayne/libmutter-fix/tree/main.
The TLDR as I understand it is that in Gnome's Mutter, when a keyboard event is received from a different source than the last keyboard event, some expensive operation is kicked off (`keymap_changed`) which is unnecessary for most cases but required to handle something like switching to a keyboard with a different layout than the one last used.
The Github repo linked above has more info and links to related issue trackers and discussions. The solution implemented in the repo is to checkout libmutter and just remove the line causing `keymap_changed` to be called for `XkbNewKeyboardNotify` events.
Sharing in case anyone else happens to have been banging their head against the wall about this
"When we started the log analysis, we first used frontier models behind commercial APIs. This did not work: the analysis requires submitting large volumes of real attack commands, exploit payloads, and C2 artifacts, and these requests were blocked by the providers' safety guardrails, which cannot distinguish an incident responder from an attacker. We ran the forensic analysis instead on GLM 5.2, an open-weight model, on our own infrastructure. This had a second benefit: no attacker data, and none of the credentials it referenced, left our environment."