One thing I learned this month:
Keep going, but know when to stop for a day.
Sometimes after a rough stretch, taking a proper break is exactly what lets you come back and finally see things start working out.
Rest is not the opposite of progress. Sometimes it is part of it.
@voidyaps I like this framing a lot. Not every browser has to become the next Chrome. Sometimes the whole point is just someone making the tool they wish existed, then finding the few thousand people who wanted the same thing.
@Yuchenj_UW Kimi K3 being good enough that people just forgot to reach for Sol/Opus/Astra for a day is probably the strongest endorsement you can give an OSS model.
@scheemunai This is exactly the kind of AI use case that still feels like magic.
Not "write me a paragraph." More like "here is a real problem, go build me the tool for it."
@tobi This is the wildest part of coding right now.
Mild inconvenience -> "fine, I will just build the software" -> suddenly you have a whole product.
Still working on Tumble.
Growth has been slow, but seeing people actually discover and download it makes me want to keep improving it.
If youโre into vintage film-style photos, give it a try:
https://t.co/rFwnrtjhvQ
Cambium is live on both iOS and Android ๐ฑ
If you have plants at home, give it a try. It can identify plants, check their health, diagnose problems, and suggest what to do next.
Would genuinely love more people to try it and tell me what you think.
iOS: https://t.co/XsCpiNBmqy
Android: https://t.co/gmr8wvuPJq
On the other hand, Grok 4.7 has been a pretty big disappointment for me.
After using Opus 5.5, the contrast is even more obvious.
I really hope Grok 5 brings a major jump in intelligence without sacrificing what Grok has always been great at: speed.
That combo would be seriously hard to beat.
GOATED release from Anthropic.
Opus 5.5 is faster, cheaper, gets higher Claude Code limits, and apparently goes further within those limits too.
They basically checked every box people actually care about.
More capable is nice. More capable + faster + cheaper is the kind of release that actually changes how much you use the model.
Opus 5.5 performs at the level of Fable 5.1. It's ~30% faster and ~40% cheaper than Opus 5 per task.
In Claude Code:
- 5-hour session limits increase 20% today
- Opus 5.5 is priced lower, so it goes 25% further within limits
- Pro, Max, and Team users get a reset to use anytime
Opus 5.5 is out and the speed is kind of ridiculous.
It feels noticeably faster in actual coding workflows, not just benchmarks.
If this is where frontier models are heading, the latency gap between โAI assistantโ and โjust part of the IDEโ is disappearing fast.
@DanielSmidstrup Exactly. The scary part isnโt even that one stream slows down, itโs how quickly platforms, algorithms, and entire markets can change underneath you.
Diversifying feels less like โdoing moreโ and more like buying yourself breathing room.
@DanKornas This is a really nice direction. A provider-agnostic terminal agent feels way more useful than being locked into one model ecosystem, especially for dev workflows.