as an open-source alternative to Grammarly
intentionally avoids including any kind of generative AI in any part of our processing pipeline
Isn't that what Grammarly is all about, though? Be better than traditional spellchecking through LLM?
I assume Harper is entirely Rules based, then? Which inherently means limited to what rules where introduced manually and what the rules cover.
A Python-specific question is better suited to the !python@programming.dev community instead of the general programming one.
The official Anthropic post/announcement
Very interesting read
The math guessing game (lol), the bullshitting of "thinking out loud", being able to identify hidden (trained) biases, looking ahead when producing text, following multi-step reasoning, analyzing jailbreak prompts, analysis of antihallucination training and hallucinations
At the same time, we recognize the limitations of our current approach. Even on short, simple prompts, our method only captures a fraction of the total computation performed by Claude, and the mechanisms we do see may have some artifacts based on our tools which don't reflect what is going on in the underlying model. It currently takes a few hours of human effort to understand the circuits we see, even on prompts with only tens of words.
Without having looked into it, I find it plausible that it could take several patchsets to come to an assessment of consequences and conclusion. Especially as they happen alongside assessments and discussion. The patchset number may also be largely irrelevant depending on what was changed.
Yes, absolutely.
I've been using it for a while. I'm not super/very command-line centered, but I use it as my default shell on Windows. I did not set it up as a default shell on [my] Linux servers.
There was a bit of a time investment in making the switch, but I enjoy it a lot, and have used a few data querying and transformation functionalities that would have been much harder in other shells, requiring additional apps; I would have probably created custom C# CLI apps for them. Nushell allowed me to do those in the shell, directly, with native operations.
I've also set up a few very useful aliases and commands; Like dl for yt-dlp, and dl opus for downloading highest quality opus audio. Or ff for a few ffmpeg conversions.
I've also contributed a bit upstream. Maybe I'll get more into Rust and be able to contribute more, and to the core.
I have my setup/configuration and scripts in a public nushell-config repo.
Examples, of how I used it productively:
Download my paged Steam reviews because Steam doesn't provide GDPR compliant exports, transforming the awful shitty HTML with query web into structured data, and transforming it into Markdown files for my website.
At work; For a list of device IDs, create JSON command files, transform them into BSON via CLI call, en-mass. (We have multiple hundreds such devices. Configuration and firmware updates require mass-updates via individually addressed command files.)
Parse and analyze DMARC reports for reported issues, and to identify report format differences from different reporters.
I'm sure I did more things, but that's what came to mind right now.
Because of how much I love Nushell, I've created a community a little while ago, !nushell@programming.dev, if you're interested.
Exactly. It's a matter of barrier and interest. Signup requirements are a barrier to drive-by improvements and reports, and them as entry points to further contributions.
A strength of the GPL is that the community can fork projects, and "take them over" that way.
At the same time, and this instance is such a case, on a centralized platform, projects can be taken over instead of be forked.
They developed and published a plugin. Now it's been taken over by someone else, on the primary distribution and discovery platform, and they have no control over it. Worse than that, the takeover now offers their sold functionalities for free.
This makes the "open source but not free, but after two years true FOSS licensed" licenses look very useful if not necessary for businesses and developers that want to monetize. At the very least when they [have to] use centralized platforms.
if the app is properly tested memory issues won’t happen
yeah, no, not really
https://security.googleblog.com/2024/09/eliminating-memory-safety-vulnerabilities-Android.html
the percentage of memory safety vulnerabilities in Android dropped from 76% to 24% over 6 years as development shifted to memory safe languages
They were prevalent. Then they started with adware. Then they bundled adware without developer consent.
Eventually they were bought with a goal of deshittyfication. So theyre fine and have good offerings, but the UI and UX is much worse than other platforms (never improved).
https://en.wikipedia.org/wiki/SourceForge#Adware_controversy
I would have expected that at least all the frameworks were using the same database, but no. Some frameworks use MySQL, others use PostgreSQL. Some framework implementations are ultra-optimized and others are what you would expect in your average web application. Some frameworks are using proper templating libraries, others are using Sprintf.
OCI does the platform hosting and development. OCF does the non-profit org work. OCF pays the OCI.
At least that's what I read from it.
I wouldn't be surprised if only weeks from now involved people announced a competing/new platform without this setup.
Sharing, because I had to look up Abstract Wikipedia