Tried openscad and then cadquery for some geometry iteration projects and found them clunky. It wasn't just that I was missing a UI; the functions, constraints and geometry kernel weren't as powerful as onshape, which I've used a bit, and presumably light years behind fusion 360, which I haven't used.
Even freecad, a UI-based oss cad, is not quite ergonomic for a beginner-to-intermediate user, though it has come a long way in the past few years.
I'm excited for there to eventually be a good open source cad option, whether language-only or language-plus-GUI, but am also increasingly on team 'tools matter for your productivity'.
Heh, the bit about context engineering is palpable.
I'm writing a personal assistant which, imo, is distinct from an agent in that it has a lot of capabilities a regular agent wouldn't necessarily need such as memory, task tracking, broad solutioning capabilities, etc... I ended up writing agents that talk to other agents which have MCP prompts, resources, and tools to guide them as general problem solvers. The first agent that it hits is a supervisor that specializes in task management and as a result writes a custom context and tool selection for the react agent it tasks.
Weather comp + low load comp + PID which means your room temperature works at the precision range supported by your temperature sensor. In my case, within 0.02 Celsius. Saves energy and makes your house more comfortable. Operated via home assistant.
There are so many good concepts to borrow from in pilot training, it's almost ridiculous. I'm not even a pilot but have studied risk management, crew resource management, decision making, etc. Anecdata of course but I feel it made a pretty big difference in dealing with projects and problems.
Bertrand Meyer suggested another way to consider this that ends up in a similar place.
For concerns of code complexity and verification, code that asks a question and code that acts on the answers should be separated. Asking can be done as pure code, and if done as such, only ever needs unit tests. The doing is the imperative part, and it requires much slower tests that are much more expensive to evolve with your changing requirements and system design.
The one place this advice falls down is security - having functions that do things without verifying preconditions are exploitable, and they are easy to accidentally expose to third party code through the addition of subsequent features, even if initially they are unreachable. Sun biffed this way a couple of times with Java.
But for non crosscutting concerns this advice can also be a step toward FC/IS, both in structuring the code and acclimating devs to the paradigm. Because you can start extracting pure code sections in place.
will give you a git repo without a working set (just the contents typically in the .git directory). This allows you to create things like `foo.git` instead of `foo/.git`.
“origin” is also just the default name for the cloned remote. It could be called anything, and you can have as many remotes as you’d like. You can even namespace where you push back to the same remotes by changing fetch and push paths. At one company it was common to push back to `$user/$feature` to avoid polluting the root namespace with personal branches. It was also common to have `backup/$user` for pushing having a backup of an entire local repo.
I often add a hostname namespace when I’m working from multiple hosts and then push between them directly to another instead of going back to a central server.
For a small static site repo that has documents and server config, I have a remote like:
So I can push from my computer directly to that server, but those branches won’t overwrite the server’s branches. It acts like a reverse `git pull`, which can be useful for firewalls and other situations where my laptop wouldn’t be routable.
Not affiliated but I'm looking into converting my commuter bike with a kit from https://ebikes.ca. Seems to make more sense to turn any of the million bikes out there into an electric one using a reusable kit than buying a single purpose electric bike that will end up on the landfill once the company goes bust because no one wants to push around a bike that heavy.
Here is my personal project that tracks my Aldi grocery prices of products over time. My rule for the project is, all code must be done by the Copilot agent within VSCode. I did hand type the Git commits.
Unifi G4 Doorbell Pro [0] is a great self-hosted option. I've been very happy with mine over the last year, but I was already bought into the unifi ecosystem with a UDM Pro SE and U6 mesh APs.
I have a Samsung Tablet and Samsung's version for said tablet is a giant mountain of crap, full of bloatware, so I installed LineageOS on it. Also my old phone and my old old phone run LineageOS because I'm just logged in to Google on my {current_phone}.
For "fun and interesting" consider an LG WebOS TV. Many can be rooted[1] which allows installing a homebrew channel[2] of unauthorized apps or writing your own.
I initially did it for Jellyfin before they made it into the official app store, but the Moonlight game streaming app has unlocked many hours of entertainment.
> While this is really handy, it is a best practice to allow users to manually toggle the color-scheme as well. Some people prefer a dark system theme, but light website themes, and vice-versa.
It can make sense for a theme selector that works on the server, since you can serve specific HTML when building the page. However, if using a JavaScript-based solution that fetches the theme preference from localStorage, I find this almost always results in a "flashbang" in dark mode, as the retrieval is slower than the first browser paint.
I've been implementing (and recommending) pure CSS-based theming to avoid this problem. Users seem much happier with them, and I haven't heard anybody ask for a theme switcher. We just respect their existing preference. However, I can see this being a problem if you offer multiple color schemes (beyond just light and dark).
I'd be curious to know if anybody has found a way to avoid this issue with JS switchers -- ideally without needing to delay the initial paint.
I do think an interesting approach would be a browser extension that lets you override the prefers-color-scheme property on a per-domain basis, similar to the toggle in dev tools.
And its lesser known component, the mailbox server used for signaling to connect the two computers. If you’ve ever installed and used magic wormhole, you’ve likely used the default public mailbox server unless you configured and set up your own.
WeasyPrint works really well for me. It can support all of the languages and fonts I need. I run it on AWS Lambda and in Docker as a web service.
I previously used WKHTMLTOPDF, but it hasn't been supported for years and doesn't support the latest CSS, etc. It does support JS if you need it, but I'd probably look at headless Chromium or another solution for JS if needed.
Looking at the prompts op has shared, I'd recommend more aggressively managing/trimming the context. In general you don't give the agent a new task without /clearing the context before. This will enable the agent to be more focused on the new task, and decrease its bias (if eg. reviewing changes it has made previously).
The overall approach I now have for medium sized task is roughly:
- Ask the agent to research a particular area of the codebase that is relevant to the task at hand, listing all relevant/important files, functions, and putting all of this in a "research.md" markdown file.
- Clear the context window
- Ask the agent to put together a project plan, informed by the previously generated markdown file. Store that project plan in a new "project.md" markdown file. Depending on complexity I'll generally do multiple revs of this.
- Clear the context window
- Ask the agent to create a step by step implementation plan, leveraging the previously generated research & project files, put that in a plan.md file.
- Clear the context window
- While there are unfinished steps in plan.md:
-- While the current step needs more work
--- Ask the agent to work on the current step
--- Clear the context window
--- Ask the agent to review the changes
--- Clear the context window
-- Ask the agent to update the plan with their changes and make a commit
-- Clear the context window
I also recommend to have specialized sub agents for each of those phases (research, architecture, planning, implementation, review). Less so in terms of telling the agent what to do, but as a way to add guardrails and structure to the way they synthesize/serialize back to markdown.
I've been running stuff in LXC for ages (and before that, custom chroots). A while ago I made the switch to Wayland - and now started moving things over to podman, which has the added benefit of being able to share the stuff easily:
I use two different setups - on some systems I only run things like browsers in conatainers, on others I also run the desktop itself in a container. Not published yet are my helper scripts, that'll need some more cleaning up.
Gluetun is the bomb. I run it in front of all of the *arrs as well as LSIO's webtop for quick access to a VPN (Mullvad). The nice thing is the operational containers don't have network access to the Internet if the Glutun container is down, so it's a nice failsafe mechanism to guarantee the VPN path.
Even freecad, a UI-based oss cad, is not quite ergonomic for a beginner-to-intermediate user, though it has come a long way in the past few years.
I'm excited for there to eventually be a good open source cad option, whether language-only or language-plus-GUI, but am also increasingly on team 'tools matter for your productivity'.