Joined July 2009
303 Photos and videos
Pinned Tweet
Apr 10
RankedAGI is alive again after way too long. I've rebuilt the entire foundation. Ditched Supabase for a custom JSON-based data layer with a dedicated admin system. What used to take hours now takes minutes. The backend friction that was killing my update velocity? Gone.
1
1
2
350
I think to deal with the recent issues and outages @WisprFlow have changed it's STT model to a much dumber version. This one is misinterpreting so many words. If yes, then Congrats! you've now changed your app from unusable to unusable, but for different reasons.
2
129
I take that back. It's now dumb, still slow and unreliable. Also the team never replies here, like they're not on this platform. They have managed to turn me from a super fan to a frustrated user, and the main contributor is lack of communication. @tankots
2
78
When Svelte with Video? πŸ˜„
Commentary is one of the most important pillars of X. And sometimes the best way to share your thoughts is with video. Today we're launching a whole new way to make them: React with Video Tap the repost button and start recording with green screen, split screen, or picture-in-picture. Now available on iOS
1
70
No other code editor makes me feel as calm as @athasdev does.
1
8
2,780
May 28
Opus 4.8 loves to say "Smoking Gun"
2
39
Tavlean retweeted
Claude Opus 4.8 added on RankedAGI State of the art in a lot of benchmarks
May 28
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors. Available today at the same price.
1
1
116
May 24
Skeumorphic slideshow controls for a custom sneaker product page that I've been working on for a beloved client.
1
2
93
May 24
Here is the screen recording showing all inner-shadow and drop-shadow values, in case anyone is curious or wants to recreate. Color palette is tailwind gray
119
Tavlean retweeted
Been using @sveltejs for a few projects lately, it's quite a nice alternative to React, fewer gotchas and complexity and Codex handles it really well. svelte.dev/
123
74
1,447
139,125
May 14
Either this is parody, or I'm blind. You don't need to be a photographer to tell how significantly worse each of the "AI Camera Assistant" version is. 🀦🏻
The new AI Camera Assistant* with Xperia Intelligence brings stories to life. Using subject, scene and weather, it suggests expressive options with adjustments of colour, exposure, bokeh, and lens for breathtaking photos*. sony.co.jp/en/xperia-1m8/s/ #SonyXperia #Xperia1VIII
121
May 14
Cursor: Pay $20/$200 get $20/$200 worth of total inference Claude: Pay $20/$200 get $20/$200 worth of inference for your openclaws, hermes, etc. And get hundreds to thousands of dollar worth of inference for your code/chat/cowork. Codex: Pay $20/$200 get hundreds to thousands of dollar worth of inference for anything that you like. One of these is close to reality, two of these are bubbles with extreme subsidies. Such wrong expectations have been set that even with the extreme subsidies the entitled devs are unhappy and want more 🀯
Claude Code weekly limits are increasing 50%, now through July 13. Live now for all Pro, Max, Team, and seat-based Enterprise users.
2
125
May 11
I just created a scene that I wished to see since I was a child: "All homo species in a group-photo wearing modern clothes" Model: ChatGPT Images 2.0 More variations and prompt πŸ‘‡πŸ»
1
1
127
May 11
Timeline infographic:
1
59
May 11
With lineage and relationship branches
28
The Claude Code max plan became an even better vfm with this update.
We’ve agreed to a partnership with @SpaceX that will substantially increase our compute capacity. This, along with our other recent compute deals, means that we’ve been able to increase our usage limits for Claude Code and the Claude API.
1
61
Something to pay attention to
Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.
55
When "smart" people say something so dumb 🀦🏻
May 2
AIs replace UIs and APIs.
1
68
Apr 18
Technically true, but misleading in a design sense. If you increased the radius to make the squircle look right in Chrome, Safari/Firefox will show a much rounder plain border-radius fallback than you intended. using @ supports and custom radius is necessary
`corner-shape: squircle;` is a very easy progressive enhancement if you want squircles with rounded corners as a fallback (the default today in unsupported browsers)
1
3
489