what kind of over-the-mars comms strategy is this?
i'm just confused over the ai era man
Starting September 14, we're permanently raising standard weekly limits in Claude Code by 25% for Pro, Max, Team, and seat-based Enterprise plans. Until then, the current 50% increase will be in place.
Readers added context they thought people might want to know
While standard base limits are increasing by 25%, this change represents a net 17% reduction in weekly limits compared to the active 50% promotional limits users currently experience, as confirmed by the official Claude account: x.com/claudedevs/sta…
6
613
excuse me? what kind of over-the-mars comm strategy is this??
Replying to @ClaudeDevs
Compared to today, this works out to a 17% reduction in weekly limits on Claude Code. We’re working on exciting changes that will make it feel like you’re getting more from Claude, while having more visibility and control of your usage. Can’t wait to share them.
2
287
shenzhen is literally living in the future.
you can book a helicopter on wechat, order bubble tea by drone, ride business on metro, get wireless charging on the street.
with the whole supply chain next door, it’s already a hub for physical AI.
what a time to live in china.
if this is true, google will likely stay behind the competition for at least 2-3 months. (provided gemini4 performance will be on par with gpt6 or opus 5.5
still remember the hype when gemini2.5 pro came out...
now do people still use nanobanana anymore?
“Industry chatter suggests the upcoming Gemini 3.5 Pro is roughly Opus 4.5 level.”
“Since our institutional note, Google has silently canceled Gemini 3.5 Pro.”
SemiAnalysis is reporting that Google’s unreleased flagship was only around the level of Opus 4.5, (a model released last year), and has now been canceled entirely.
Google reportedly shipped Gemini 3.6 Flash as a bridge, while redirecting the hype toward Gemini 4, but SemiAnalysis says it still trails Muse Spark 1.2, Grok 4.5, and tier one Chinese open models, leaving Gemini around 8th or 9th place.
Remember I posted a while ago that Fable 5 scores 80.3% on SWE-bench Pro, the much harder variant. Gemini 3.1 Pro scores 80.6% on Google’s SWE-bench Verified run and just 78.8% independently. Fable is matching / beating Google’s best Pro model while being tested on the harder version of the benchmark. So it makes sense that they’re still below fable in terms of coding. Still hard to believe Anthropic had this model back in February, one month after Gemini 3.1 pros release.
1
6
1,713
iOS and android.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
4
8
500
Talent and compute are the scarcest resources of the AI era.
But unlike compute, talent cannot simply be produced and owned.
It’s a choice of where to study, work and stay.
Félicitations à Hong Wang, lauréate de la Médaille Fields !
Formée à l’École polytechnique, chercheuse à l’Institut des hautes études scientifiques, de Paris-Saclay, elle incarne l’excellence de notre recherche et de notre formation. Fierté.
Choose France for science!
1
8
376
I gave Fable 5, GPT-5.6 Sol and Kimi K3 the same task: show me how The Odyssey feels across different screen formats, from 1.43 to 2.39.
-Fable made a guide.
-GPT made an editorial.
-Kimi made a playground.
Kimi’s version is my favourite, though it also took the longest.
Since I probably won’t get the full IMAX 70mm experience anytime soon, I wanted something that could at least help me feel how the different formats change the viewing experience.
Here’s what they came up with.
One caveat on Kimi K3:
New subscriptions were paused when I ran this, so I accessed K3 through Hermes Agent instead.
Around 10 of the 58 minutes disappeared into a detour where the agent tried to use xAI OAuth to gather more material before I stopped it.
So the timing isn’t directly comparable.
Still, the result made me curious enough that I’ll probably try Kimi directly once access opens again.
The capability looks real. Whether the service can scale with demand feels like a separate question.
1
2
42
Conclusion:
Clearly, all 3 could build the thing amazingly well. It will eventually come down to some personal judgments:
what to emphasize, what's the wow factor, and when to stop.
I’m tempted to clean up Kimi’s version a little:
– keep the frame and crop control inside one viewport – remove most of the editorial material – fix the shot timestamps – fix sound issue when buttons pressed
Interested to see how far can I take this to.
2
24
Can’t believe there is not a single IMAX 70mm screen in Asia where you can see the Odyssey as it originally filmed.
Closest thing is IMAX GT with 1.43:1 ratio (no more than a dozen in Asia)
1.9:1 IMAX screens won’t make the cut. 2.39 may even present some scenes in more classic ways.
1
15
1,318
You can see the difference across various formats on the movie’s official site.
odysseymovie.com/explore-for…
4
80
models are becoming gadgets.
model releases now get marketed like phones and laptops.
influencers and benchmarkers get early access with guided narratives. they make their content, then everything drops at once on launch day.
these reviews are basically unboxing videos.
when the marketing playbook converges, the commoditization has too.
5
404
Thinking of redo this one with Kimi K3
Replying to @adamy_ai
I gave the same brief to Claude Code (Fable 5) and ChatGPT Work (GPT-5.6 Sol), testing several visual styles with each.
In the end, I used GPT-5.6 Sol for the two-view set that came closest to the Potala I wanted to remember and share: one looking inside (post 1), and one showing how the palace sits in its landscape.
crazy
Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5.
This is a 17-place jump from Kimi-k2.6 (#18 -> #1).
In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5.
The full model weights will be released by July 27.
Congrats to the @Kimi_Moonshot team on this major milestone!
4
184














