Aaron Miller

⚠️🔧⌨️🔥 subliminal, liminal, superliminal

[Aaron Miller @apage43]

@Bitwig can _import_ tempo automation from a .DAWProject, but not _export_ it?!

[Lisan al Gaib @scaling01]

Fable 5 refused 200 out of 200 ProgramBench tasks lmao

[Aaron Miller @apage43]

@thdxr i figured the real signal that it was bigger was that it was significantly slower, but you never know.

[Aaron Miller @apage43]

@Westoncb I have tried this on a few harnesses and unsurprisingly a lot of them limit subagent depth, which makes it not work as well - Cursor desktop will do 3 deep, but Cursor cloud will not.

Also limited in that new work only starts after the current slice ends.

[Aaron Miller @apage43]

@Westoncb i just have that one running all day and do planning in separate threads which attaches work to the tracker

[Mark Otto @mdo]

Ayyy go have some fun with https://t.co/WrAtEqXtmJ, a crispy clean demo app showing off our new CodeView component for rendering millions of lines of changes with all the fixings—virtualization, streaming, scrolling, annotating, and more. Just replace `github` with `diffshub`.

[Mitchell Hashimoto @mitchellh]

This is why PR diff speed matters. This isn't a dunk on GitHub specifically, because GitLab, Forgejo, etc. are all equal or worse. But this is the kind of thing that drives me nuts, because this is a core workflow and its slow enough I literally take my hands off the keyboard.

Btw, when my mouse jiggles on the left, its because the page is literally skipping frames and I'm instinctively shaking my mouse to see if it'll respond. And on the keyboard input you can literally here me finish typing before a letter even shows up.

For someone like me who is an expert at these tools, my brain navigates the tool dramatically faster than it can keep up, and that is not good. The tool should not get in the way.

[Aaron Miller @apage43]

@garybernhardt context is precious - try to only have rules in context when they matter - i try to break tasks into <1 context window pieces each with a task-specific context-gathering plan, and have a 'orchestrator' thread that i ask to run those in subagents and clean up after them

[Gary Bernhardt @garybernhardt]

Are there any actual success stories for "software factories" like gastown? Anything I can see running, and preferably see code?

[Nomic @nomic_ai]

Today, we're launching AEC-Bench — the first open, multimodal agent benchmark for construction.

196 tasks across real construction documents. Full agent harness. Automated evaluation. Apache 2.0.

We benchmarked Claude Code, Codex, and our own agent. Here's what we found 🧵

[Aaron Miller @apage43]

@dangitman50 @ContrarianSense @Adityapandeydev can also just run it in a transaction and roll it back instead of committing if it did the wrong thing

[Aaron Miller @apage43]

@honour_can_code @gtx360ti most services will either not serve the video or serve lower quality video to a browser that doesn't support the DRM - and when they're working the DRM systems pass the frames directly to the video card bypassing the layers that capture/screenshot software read from

[Aaron Miller @apage43]

@brian_lovin that is, a given keypress always sends the same bytes to the pty device and the kernel uses the device's current flags (raw mode is just all flags off) to decide whether that has any special behavior, whether to do local echo and so on, which is why this all still works over ssh

[Aaron Miller @apage43]

@brian_lovin nitpicks: EOF isn't a signal (Ctrl-D just flushes the current line and flushing an empty line causes apps to read() 0 bytes)

also signals and line editing aren't handled by the terminal program - that's all part of the kernel's pty layer, and the shell sets what is "foreground"

[Zack Korman @ZackKorman]

If I were a threat actor exfiltrating a ton of company data I’d set my user agent to “ChatGPT”

[Aaron Miller @apage43]

@YoungMOOBY @HSVSphere @bee_fumo I think older bitlocker might actually have let you do that - I remember it also used to let you make recovery thumb drives. But yeah the current defaults are bad and get people in trouble

[Aaron Miller @apage43]

@HSVSphere @bee_fumo Feature not a bug - you're supposed to turn off bitlocker before doing bios updates

[Aaron Miller @apage43]

@HSVSphere @bee_fumo More likely because the disk encryption key is sealed against some PCRs (TPM side accumulated hash of firmware+bootloader) as protection against bootkits - may be able to downgrade to the same old version to get back in

[Aaron Miller @apage43]

@abacaj its to be a devkit for the CUDA+ARM+unified memory platform to get more software to support the datacenter versions of it