oof this is so real
I'm so grateful for my current job ๐
oof this is so real
I'm so grateful for my current job ๐
the biggest problem for me is I feel like I'm losing my community
no one gives a shit about the code anymore
I can't even talk to people about context engineering or workflows
they don't care
if they care about anything at all, it's "how can I go even faster?" or "20 agents in parallel!!!" ๐ฌ
are you aware of this?
github.com/anomalyco/op...
I can see you've been assigned
both Kimi K2.5 and MiniMax M2.5 are affected by this, it seems
there are many reports, on Reddit too
I would subscribe for these no doubt, if they worked reliably ๐
nice, Twigl is open source ๐
I've opened an issue
github.com/markusfisch/...
I'm not sure the project is really maintained anymore though...
man these are so cool ๐
I wish I could use these as wallpaper on my phone ๐
I tried pasting into this app:
github.com/markusfisch/...
but shaders rarely work there, usually an undeclared identifier error
I don't really know shaders, but the code always looks to me like something is missing? ๐คทโโ๏ธ
AI generated web sites: ๐
AI generated content: ๐
and that other medal he didn't win
item 2 is not "okay"
LLMs making autonomous kill decisions is about as dystopian as tech gets ๐
Open models are over just as fast as they arrive.
I've tried two of these (Minimax M2.5, Kimi K2.5) but I think I'll just skip GLM5 and withdraw from the pursuit of open models for another 3-6 months.
There is real progress, but it's slow, and it's never what YouTubers promise every single day. ๐
it goes through the full RGB spectrum, just like a regular RGB rainbow
an OKLCH rainbow goes through all the hues but not at full saturation or lightness
speaking of
I've been mildly obsessing about the UX of hue pickers lately, and finally came up with a solution tonight
github.com/omgovich/rea...
it's the full RGB spectrum, but the hues are distributed according to their OKLCH based contrast ๐ค
feel free to steal the idea ๐โโ๏ธ
terrifying (and appropriate)
I asked Claude, and it called this "a reliable signal for content worth skipping" ๐
why is "100% Prettier compatibility" always the first thing advertised by every alternative to Prettier?
Prettier is awful - cramming as much code into every line as possible is by no means any sort of "gold standard" that formatters should chase ๐ฎโ๐จ
please try dPrint!
and spread the word ๐
yeah the debate lacks nuance
every argument of this type goes nowhere, just leads to more ๐ฅ
I'm not even sure it makes sense to debate these matters on the short format messaging platforms... the context window is too small ๐คช
looks like JPEG-XL support is once again under way in Chromium ๐คฉ
issues.chromium.org/issues/46291...
if you recall, it was removed over politics and disputes over patent grabs by Microsoft and Google ๐
Microsoft has already commented and asked that the feature be made optional ๐คฆโโ๏ธ
hahaha, shipped where, Zach? ๐
a logo that looks like a placeholder for a logo that failed to load?
not really, right? seriously? no.
you might see nothing new here, but the fact that he *stays in flow* ๐
I've never seen anybody work with AI in this way - most people wait for the AI to complete a step, then review.
this feels more like coding, accelerated ๐ค
youtu.be/ws9zR-UzwTE
they'll write this off as "a bug", but it's just how these models work: the base model is essentially a form of lossy compression - everything they stole is in there... or some sad, mangled form of it. ๐
calling it:
@penpotapp will eat Figma ๐
github.com/penpot/penpo...
this isn't 100% yet, but it's already close
and the design tokens system in PenPot is so much more powerful than the mess of variables and styles in Figma
incredible progress! I am so excited for this ๐
also they should be styled to look like radio buttons, or users won't intuitively understand that the options are mutually exclusive ๐
have you tried PGLite though? it's an embedded version of Postgres
full text and vector (!) index might really open some interesting avenues
what is this new BS where tablets are advertised as "24 GB (8+16 GB RAM)"
these devices have 8 GB of RAM and 16 GB of reserved swap space on SSD
this isn't even a new feature, they've just now discovered a great new marketing deception
this is wildly misleading
SSD is not suddenly RAM! ๐ฎโ๐จ
I don't think most people are saying that?
just my personal impression, but most devs I know don't think that - and most of them use OpenCode or other agents.
again, just my impression, but it really seems like those who think that just make more noise than the rest of us. ๐
Introducing FragCoord: My ultimate shader editing tool!
I think they are already?
Kimi K2.5 is #1 on openrouter rn for programming and tool calls
I tested it with the latest version of Opencode yesterday and I'm really impressed
Google, Anthropic or OpenAI models could not have done this task better ๐
that's normal, none of us know why we're here, we just know we don't want to be at the other place
nice ๐
this talk has some good points as well:
youtu.be/DZpR0GojoWQ
"if you don't know how it works, then you don't know that it works" ๐
they don't pay us to just churn out code, we're paid to take ownership and responsibility ๐
cooperative multitasking is the only reason Node works at all, right? any excuse I mean reason (ahem) to cycle the event loop ๐