• 0 Posts
  • 8 Comments
Joined 3 年前
cake
Cake day: 2023年7月2日

help-circle
  • Forgive me

    No. Almost all of this response is laughable, and has the same stench as the project. I was polite, I wish well-meaning projects well, and if it weren’t for the absurd evangelism that someone else might trip over and be swayed by, I’d just leave this be.

    These are not high-profile developers. We don’t know their energy or their experience. We owe them nothing. And whatever this is

    If anything, this is the most legit and inspiring use case for AI I’ve seen in a while.

    is completely ridiculous.

    And for people saying that they are just using the AI products of the companies that dropped them… again they must be far from the industry. There’s a small handful of companies building the leading models and those are not the companies that laid these people off. There is a huge world of SAAS companies like Altlassian and Salesforce and thousands more you’ve never heard of who build business productivity software - bloated, shitty, overpriced crap that’s ripe for displacement. Nothing to do with LLM vendors.

    Take this meandering crap back to the bot and ask it to narrow it in on an actual point. There is no path for Claude to liberate anyone from platform capture. It’s intrinsic to the platform Claude itself. CRM will be free from Salesforce, and we’ll still owe Anthropic for the tokens. Perhaps a noble goal, but there’s no freedom in trading one master for another.


  • I worry that every level has this misunderstanding, that “Salesforce” is a monolithic system that can be remade if only we pass a comprehensive test suite. Salesforce does not have 80k+ developers making a singular monolithic product. They have 80k+ employees servicing the sales demands of a massive number of customers. Excel is a perfectly fine CRM if you only have a handful of customers.

    Hope you’re staying sane in the SaaS landscape. I hear a lot of horror stories.






  • That simply is not what people want when they look for information.

    Well, except for those who do. The problem is a use case mismatch. I’d argue, if anything, an encyclopedia should contain the minutiae. Unfortunately, there’s no huge compendium of brief but accurate and sourced synopsis of the same topics. To be fair, we’ve never really had one.

    I agree with the editors that embedded AI summaries are not a good idea (at the moment, at least). Users can bring summarizers to the data set of that’s their want, or someone (maybe even wikimedia) will find a way to provide this in a way that preserves the underlying data’s validity. Stripping Wikipedia of its full context seems like a bad idea.


  • Headscale, for one. This is probably implied as part of one of your above stacks, but let’s list it out loud. Tailscale is great and all, but it’s downright icky to offload routing of any variety to a third party.

    Immich. Turn off Apple or Google’s automatic scraping of all photos, keep usability. Even if you’re not a photo person, at least some of your users are.

    Syncthing is or Nextcloud, or something in the family. This may already be part of your NAS plans.

    One of the code forges like forgejo, gitea, gitlab. Even when not a developer. Self hosting involves configuration and if you can get that into text and into a history, it makes things so much easier. Add bells and whistles to your hearts content, but these are good suites for a lot of functionality. Forgejo does have federation on its road map, but it’s a while off still.

    These are ones I find pretty ubiquitous. There’s so many options once you have initial infrastructure. Email, for instance, isn’t as daunting as the horror stories make it sound, though not as simple as many hope. My suggestion is to take time and do it correctly. There’s a lot of backtracking involved as you learn more, but it’s usually worth it. Best of luck!