“Most software that requires hiring and paying software engineers has low risk tolerance”
Is this true? I’ve worked in various software companies for over 20 years now and I’ve never had to worry about risk in particular, and the codebases were all somewhat bad in areas and buggy (as tends to happen when a codebase gets large and old).
The past 6 months LLMs have crossed into barely needing to check the output territory for what I work on, and generally write very similar code to what I was going to write. It needs some common sense to use it correctly of course, like going feature by feature and keeping the commits fairly small, but my job is easily 10x less work/time for the same results.
I’d be genuinely surprised if even 1% of software requires aviation levels of risk tolerance and testing, but maybe I’m mistaken and most people are working on much more mission critical things than I am?
Can you fillet unions? Like, if I union a cube with a smaller cube sticking out of its side, can I make a fillet so that the edge where the two cubes' faces intersect doesn't become a stress concentration point for cracks to start at? This is OpenSCAD's primary weakness from my point of view.
Not the one you're replying to, but "my" system is an Angular application that makes heavy use of CodeMirror.
One instance for the base editor, and one instance for the currently active cell. The other cells are rendered to HTML through a different code path (no CodeMirror involved there).
For rendering mathematical formulas it uses MathJax.
A global context-based keyboard handler such as the one in VSCode allows for Emacs key bindings.
I've vibe coded a codemirror table editor (as a plugin for joplin) using a nested codemirror instance for the cell editor, in my case using joplin's built in command to render markup for the inactive cells: https://github.com/bwat47/joplin-rich-tables
it works well but I'm not much of a programmer and probably ended up with a crazy over engineered architecture lol
Sure! At first I wanted a sublime that played nice on windows with virtual desktops (it has some annoyances when dragging tabs, they dont respect separate destkops). I use sublime as my 'ephemeral' text editor, and Obsidian for my actual saved/persisted notes, but i love Obsidian's WYSIWYG markdown experience so i wanted to bake that into this as well.
So the core ethos of it is: 1. ephemeral AND safe - easy to jot notes but also safe from crash or accidental closing ( all text buffers persist a rewindable history), 2. performance, 3. minimalism: I don't want a whole plugin ecosystem or bloat. extra features are all accessed via the command palette, 4. markdown WYSIWYG rendering tailored to my liking
Technical details:
Native markdown notes editor for Windows, written in Rust. No GUI library as I went straight to raw Win32 with DirectWrite/Direct2D/D3D11 underneath. No winit, no wgpu, no async. Just a real wndproc and threads talking over crossbeam channels. The rope is canonical and there's a separate "display map" layer that handles wrapping, folding, hide/replace projections.
source bytes never touch the GPU and every text layout is built from a display string with styles baked in. SQLite is the source of truth and the files on disk are exports, so every keystroke is durable
Performance is the whole point as I wanted sublime-level performance(and beyond) There are gated budgets in CI for every slice, eg. keystroke to pixel under 8ms p99, decoration parse under 1ms, D2D submission under 2ms, edit-to-durable under 400ms, plus a variance-tail contract (p99.9 ≤ 2× p99) so sporadic stalls fail the build too. Steady-state target is zero new heap allocations per keystroke. Layout caches are keyed so they survive typing bursts, the display map has splice + motion-reuse caches that killed O(document) rebuilds on single-character edits and caret moves, and every reflow funnels through a helper that pins the caret's screen-y so font/wrap/resize changes never make you re-find your cursor.
I've also done a bit of research on the psychology of perception and response times to try to tune this for maximum intuitiveness and responsiveness. Too fast is disconcerting, too slow is infuriating, too little feedback is confusing, too much is distractinc, etc.
this has taken about a week so far and it's almost ready to replace my usage of sublime
Nice, thanks for the link. It actually works, but lacks basic features as opening the whole folder with subfolders, pressing spacebar to play or loading large amounts of files, the app hangs immediately.
Amazing, thank you for the reamp tip! I used winamp always in my Windows years like 15yrs ago. But then I switched to Mac and forgot about winamp completely.
I’ve run into this exact situation several times recently. They were all tasks that I’m positive Claude Code could have managed in the past, but now it simply cannot get over the finish line a lot of the time, and Codex will one shot fix it just like the old Claude Code could. I’ve even tried having Code fix and implement some old tasks it had done correctly in the past and now it simply can’t.
My guess is that it is the fault of the model rather than the harness, I believe Opus to be much worse than it was for whatever reasons. Though I suppose it could be Code’s fault somehow. For the time being though Codex is much better which I never thought I’d be saying.
I plan to run tests using Pi so they have the same system prompt and harness, but I’m suspicious that it’s only the subscription level Claude Code that is worse and we’re not allowed to use that with Pi.
Agreed, it's very strange. I'm sure there are many projects that are like they describe, but it's certainly not all of them. I have worked as a game dev for over 20 years, and probably 75% of that time my team and I have been coding. AI has been an incredible game changer for me over the past 6 months or so (I was using it quite a bit before then, but the capability became much higher lately). I actually have some free time in my days now while still hitting milestone dates, instead of endless crunching.
Is this true? I’ve worked in various software companies for over 20 years now and I’ve never had to worry about risk in particular, and the codebases were all somewhat bad in areas and buggy (as tends to happen when a codebase gets large and old).
The past 6 months LLMs have crossed into barely needing to check the output territory for what I work on, and generally write very similar code to what I was going to write. It needs some common sense to use it correctly of course, like going feature by feature and keeping the commits fairly small, but my job is easily 10x less work/time for the same results.
I’d be genuinely surprised if even 1% of software requires aviation levels of risk tolerance and testing, but maybe I’m mistaken and most people are working on much more mission critical things than I am?