george stephanis toots on mastodon

Brightly haired wierdo. Neurodivergent. Technomancer. However old being born in 1984 makes me. He/Him. 👉 Trying to move to [email protected]

[george stephanis toots on mastodon @daljo628]

@joinmastodon oh and I'm over here: https://t.co/Ztf0GGp2aF

[george stephanis toots on mastodon @daljo628]

From what I recall in previous months, the big ones were priced at 1x instead of the current 6-9x. I wonder if the rates drop in off-peak hours as well.

[george stephanis toots on mastodon @daljo628]

Also, for what it's worth, I think a lot of the people who just ran out of their requests this month on Copilot in like ... a day, may not have realized how the model pricing has shifted.

[george stephanis toots on mastodon @daljo628]

Its default options? Now they're priced at 6x and 9x when I look. So if you leave it on "Auto" it'll give you a 10% discount of the jacked up rates.

For smaller stuff? Pick a smaller or older model! It'll run for a fraction of the cost. Or play with local models.

[george stephanis toots on mastodon @daljo628]

@SybreWaaijer That's how I got into core contributing fifteen years ago.

[Simon Dickson @simond]

Just landed in WordPress plugin repo: something I've been working on for months, and have had in mind for 10+ years.

Imagine never having to ask a client what they mean by 'broken'. Imagine never having to onboard them to a project management platform.

https://t.co/Moj7DyTcUU

[george stephanis toots on mastodon @daljo628]

@austinginder Okay, but here's the really fun part: When you have the agent do this, have them document all the steps and gotchas and the like to turn it into a reusable skill. :)

[george stephanis toots on mastodon @daljo628]

@famousish @OpenAI Any that you're comfortable being public, publish in a github repo for others to use?

https://t.co/JXPbZVNvtd

[george stephanis toots on mastodon @daljo628]

@famousish ... please be running backups in case it goes off the rails

[george stephanis toots on mastodon @daljo628]

Also, if you're suspicious of LLMs ability to translate between languages? Yeah, same. But it's 1) a cool demo of the technology, 2) better than nothing, and 3) easy to manually edit and correct afterwards.

[george stephanis toots on mastodon @daljo628]

Wanna see some of the cool stuff WordPress 7.0's LLM Connectors can do?

https://t.co/fKedVshV1q

I built this for something else, but realized it'd be a fun proof of concept to use them to quickly generate translations for bits of content.

[george stephanis toots on mastodon @daljo628]

I've been working with vLLM for local models, tying WordPress in with those instead of the big commercial hosts -- but Ollama or LM Studio are both great options as well.

WordPress Playground preview link's available on the repo to give it a try!

[george stephanis toots on mastodon @daljo628]

@famousish Want a blast from the past?

I still use this wallpaper.

[george stephanis toots on mastodon @daljo628]

@film_girl oof inflation hits hard. I'd been yearly for a couple years then sprung for lifetime in 2021 for $119 + tax.

I wonder if the backlog of no-longer-paying lifetime pass holders means they need to lean on new subscribers to pay for the service, which has to jack up the price?

[george stephanis toots on mastodon @daljo628]

@austinginder I'm pretty sure there's some tooling already in place from the Tide project -- https://t.co/YZ62qRw0OA -- cc: @jeffpaul who I think could at least get you in touch with the right folks?

[george stephanis toots on mastodon @daljo628]

@austinginder I mean, you should be able to pull that data from the plugins svn repo firehose?

[george stephanis toots on mastodon @daljo628]

@austinginder @mholt6 My stance from when I used to review hackerone reports is that I don't mind automated scanner reports, so long as they're 1) significant, and 2) verified by a human before submitting with a proof of concept.

[george stephanis toots on mastodon @daljo628]

@courtneyr_dev Do you have any idea how many client sites I work on that do this? I've taken to just silently fixing it for them.

[george stephanis toots on mastodon @daljo628]

Dear humans who make wordpress themes: If you're going to `position:sticky` your header nav in WordPress, please also give it `top: var( --wp-admin--admin-bar--height, 0px );` -- so it doesn't render /underneath/ the wp adminbar.

<3

[george stephanis toots on mastodon @daljo628]

@famousish @3Utsav Hit me up in slack if you wanna bounce ideas back and forth as you start trying, or want some resources.

[george stephanis toots on mastodon @daljo628]

@3Utsav @famousish Genuinely, I've been playing with local models as well. It's less "magic" as you have to do the setup, but claude is great at advising on configurations. Ollama's much easier to set up, but vLLM does better with performance.