shibel

Thanks I just want my GitHub actions to not break every so often though.

I disagree with the first paragraph, agree with the second, and am not very hopeful about the third.

Love it (the mockery, that is). So far Opus 5.5 feels noticeably better in this regard (oh, and it’s fast!), but I felt the same about Fable initially. Not buying into the nerfing theories yet but if I find myself going into LLM-tilt again soon then I might join the club.

I don’t care about pricing, I don’t care about speed, I don’t care about the agentic coding improvements. Those are already fine. Does it still reply with walls of invented jargon, stitched-up phrases, and manage to cram 10 concepts/subjects in one sentence?

Your comment doesn’t contradict mine and mine doesn’t contradict yours. Linear’s MCP is just (very?) inefficient. I explain the source of the difference in the last two paragraphs.

Funny you mention Linear’s specifically, I just posted about moving to a CLI instead because of the MCP’s egregious token-usage [1]. While this doesn’t discount the points you mentioned, I think the context savings (which can be huge, I hadn’t listed all differences in that post) outweigh them specifically in Linear’s case. It is just too inefficient in that regard.

Edit: this of course says nothing about MCP vs API/CLI in general. It’s just a bad implementation by Linear.

1: https://thebiglog.com/links/linear-cli-instead-of-linear-mcp

Yes, that’s my biggest gripe with Claude at the moment: https://news.ycombinator.com/item?id=48789551

It’s more inhibiting to me than bad code, missing tests, or an obvious or non-obvious suboptimal solution.

I’m (sorry for the lack of humbleness) a very fluent non-native speaker and writer, and this is by far my biggest challenge with Claude. It stitches together 2-4 advanced concepts into one or two words and I always have to ask it to “unpack”.

I don’t think it’s easy on native speakers when it happens, but it’s even harder when you’re not.

This is crazy. I just got the CCX33 last week to migrate a Heroku db to it, that seems to have jumped over 100% now…? Need to reconsider

Edit: just noticed this is not retroactive. Still concerning looking forward.

Super interesting. I’ve been failing hard at gaining weight. Intrigued. I hope you don’t hesitate to make the app paid if you find the justification.

A bit O/T but:

> I have never been a big fan of pull requests.

I guess this partly explains why Zed (still) lacks a PR review flow, let alone a coherent one, despite some interest [1]. Pretty much the only reason I’m still with JetBrains.

[1]: https://github.com/zed-industries/zed/discussions/34759

Maybe…I really didn’t want to have to install another app just because I’m sharing though. My line of thinking was/is:

If tomorrow I decide not to share with anyone, I don’t want to have to reconfigure stuff. I simply edit the Tailscale policy file, and (maybe) spin down my server.

I need to finish off that blog post.

With Tailscale, you can basically point a domain to the FQDN of a machine you’re sharing with people and the domain will simply work for them (and only them)[1]. But for it to work without them having to know or specify the specific application port, you have to grant them access to 443 (and 80) in your Tailscale ACL for that specific host.

So yeah, now immich.familydomain.com works without family members having to remember the specific port. BUT, serveradmin.familydomain.com (another app on the same host) will ALSO be accessible to them (from a networking POV). We opened port 443/80 for that host after all.

I took a few hours with Claude back in January (?) and we wrote a tiny Go authorization gateway which basically consults both Tailscale’s public API and Tailscale’s `localapi` and returns the appropriate response to Caddy based on the requesting user’s actual allowed ports.

So now I can share different apps (subdomains) with different people without forfeiting access controls, all driven by Tailscale’s policy file.

(I hope I didn’t mess up the crux of it, pretty late here)

Edit: why not (something like) Authentik? Quoting from the draft:

  I contemplated this for a bit but one thing I kept thinking about is:

  _They're already logged in to Tailscale_

  Why do I need to install and configure another full-blown app dedicated to doing something Tailscale already did? Why have the users go through two hops of authentication?”

[1]: https://www.youtube.com/watch?v=Vt4PDUXB_fg

Reminds of a neighbor I had back when I was renting in a big city. He didn’t seem to understand what’s wrong with keeping his TV on for very long periods broadcasting the sleaziest (at least at the time) reality show on full volume.

I tried talking to him multiple times to no avail. He’d basically say “yeah I’ll pay attention no problem” but nothing changed for weeks.

Coincidentally at that time I was working morning shifts at a radio station. Those start really early so you gotta wake up at around 4am.

I decided one day to change my alarm (triggered on my Sony Vaio) from the peaceful iPhone-like tunes to System of a Down’s “Chop Suey”. I also decided to forget it on, on repeat, full volume, while leaving the apartment.

I don’t think 3 days passed before he knocked loudly at my door, moaning and complaining.

I told him: “you gotta understand, your TV was so loud I couldn’t sleep for nights on end, the old tune wouldn’t wake me up anymore. I had to change it. I’m so tired that I even forget to turn it off.

But yeah, I’ll try to pay attention to it”

Astro is great. It checks all of my checkboxes. I hope this is not the beginning of the end.

A custom Python/Django based mini-app (mini — at least for now) that will allow me to import transactions sanely and safely to my local GnuCash install.

I’ve been doing tedious manual entry for a bit over two years now and after having missed three consecutive months, the only other option was to bail.

As a start it should help with 3 main things:

- Translation, categorization: my source documents aren’t in English but my GnuCash entries are. This is one of the reasons I don’t use the built-in imports. (This should shave off at least 90% of time spent entering data)

- Human-error prevention: there were at least 5 times where it took me over 15 minutes to reconcile a discrepancy because I entered some number or some account wrong somewhere.

I happen to use Paddle. Happen because Stripe isn’t available in my country and Lemon Squeezy wasn’t around / was very new back then. It’s OK but it really irks me that they don’t really fight chargebacks. (I haven’t had many but the fee they throw on you automatically is so hefty for a B2C.)

Their V1 API (not Paddle “Billing”) is…not the best I’ve used.

Off topic: I’m halfway through the article and can’t help but notice the relatively high number of times a word is erroneously repeated twice; wondering if it was “edited” by AI.

Yeah, my first lines of Python (or code) were Selenium/BeautifulSoup. The problems with these are sensitivity for (DOM) changes, speed, and scale (not that I'm too concerned with those at the moment, it's just a fun refreshing project).

Yep, I don't have much experience with affiliate programs (I like to build more SaaS-like apps[1]) but from my research Amazon have gotten really stingy circa 2019. Happy to see an alternative.

[1] I didn't know about the 3-sales precondition either. I guess some people bought albums on Amazon on my other music-related app.

Yeah, I also wondered how quick would it be to ~~rip off~~ replicate “diskprices.com for X” and took a weekend (ok more) to build https://tvpricesindex.com (Also using HTMX).

TVs are much less of a “commodity” of course, but it’s a nice experiment that I’ll iterate on a bit more with time. Also, first deploy on Railway (was nice enough, definitely compared to Heroku, but they have a way to go), and using other libraries I wanted to learn.

Will probably start grouping models better soon, and offer other filters.

BTW: If you plan to do something like this with PAAPI (Product Advertising API), know that Amazon has the constant axe of banning you if you don’t generate “qualified” sales for 30 days straight.

What’s the point of having [Spoiler Alert] in the opening paragraph if the spoiler is already in the title?

1) A friend in DevOps who supposedly has contacts there.

2) It's not like this push is a big secret: https://www.cnbc.com/2022/10/12/microsoft-github-relying-mor...

Rumor has it Microsoft is pushing them to move their infrastructure to Azure, which explains much of the frequent downtimes lately.

Hi. I’m the author. First off, happy to hear you find that refreshing - it’s a really nice compliment to hear!

I, like you, hate blanket statements. Perhaps I should I’ve emphasized a newbie like me in that sentence? I really knew next to nothing about how web development, and specifically serving your site from a server, works.

Author here. Small world! Would love to see your app if you don’t mind sharing. I’m @SHxKM on Twitter if you’d like to connect.

OP here: that’s a good idea that did come up to mind (specifically rotating the Redis queue credentials). I was just sure Heroku’s gonna take care of this shit by now.

BTW, I’m not 100% sure it’ll work: if Heroku updates that zombie dyno’s environment variables, it may still have access to the new credentials :/ (My Redis instance is also a Heroku add-on)

OP here: I agree my phrasing is kind of problematic. Will rephrase shortly. Thank you for bringing that up.