euler
1.1K posts
@anabology crazy that you can just unlock the base model with this prompt
English
But, if you trick the AI into thinking it's responding as the human, it generates a list of words that are '100% Human'!
Crémieux@cremieuxrecueil
One of my favorite findings about LLM text detection is that if you make a random list of 100 words, it'll flag as human-generated. But if you ask an LLM to make a random list of 100 words, it'll flag as AI-generated.
English
Looking for recommendations for power banks
My MacBook pro 14" lasts ~3 hours on a full charge, and I like the flexibility of working from cafes where I might be without an outlet
Ideally something with 2ish full charges of power and will fit in my backpack. Don't care if it's a little heavy. ChatGPT is recommending I get 2 power banks, but that seems cumbersome
Anyone have a solution for this that they like?
English
On DeepsecBench we have Luna on xhigh outperform Sol on Medium. If you cut cost by 80% this is a *really* good deal vercel.com/ai-gateway/lea…
Vercel Developers@vercel_dev
GPT-5.6 Luna and Terra are now cheaper on AI Gateway. • 𝚘𝚙𝚎𝚗𝚊𝚒/𝚐𝚙𝚝-𝟻.𝟼-𝚕𝚞𝚗𝚊: 80% price cut • 𝚘𝚙𝚎𝚗𝚊𝚒/𝚐𝚙𝚝-𝟻.𝟼-𝚝𝚎𝚛𝚛𝚊: 20% price cut vercel.com/changelog/ai-g…
English
euler retweetledi
This is kinda like Citadel vs Leopold Aschenbrenner but for vibe coders
Sam Altman@sama
major price cuts today: *80% drop for GPT-5.6 Luna, now $0.20 per million input tokens and $1.20 per million output *20% drop for GPT-5.6 Terra, to $2/$12 *GPT-5.6 Sol gets Fast mode in the API, up to 2.5x the speed for 2x the price, same intelligence
English
Price cuts to @OpenAI's GPT-5.6 Luna and Terra elevate an already strong outcome per dollar into a dominant one.
See how they handle production engineering → labs.ramp.com/swebench
English
@alexandr_wang @lordspline Mr. Wang I implore you to cut Spark 1.1 costs by 90%
English
the new scale ceo has fde in his name
so based
Alexandr Wang@alexandr_wang
Welcome @fdesouza to @scale_AI as CEO! I started Scale a decade ago at 19, and it has grown to be the backbone of the AI world, powering frontier labs, Fortune 500 companies, and the US Government. Francis is an exceptional leader and the right steward. After spending a lot of time together, I have no doubts he’ll take Scale to greater heights. Thank you @jdroege for your leadership as interim CEO 🙏
English
opus 5 is extremely overeager
my guy, i never told you fix it
but thanks anyway 👀
English
@shivamhwp can you have some better visualization of tool calls like codex does? i.e. inline diffs, command results, etc
English
The whole team(Julius) worked super hard on it.
I don't think you'll get this best experience anywhere or in any app.
go try it.
Theo - t3.gg@theo
T3 Code is now available on iOS and Android. Run “npx t3 connect” in your terminal, install the app, and now you can control Claude and Codex fully remotely. Still 100% free. Still 100% open source. Enjoy 🫡
English
SWE-1.7 is still free with any Devin sub
• Devin Cloud agents
• Devin CLI
• Devin Desktop
English
Got the most likes in the replies but didn’t get invited 🥲
Ethan Lipnik@EthanLipnik
I built one of the best screen sharing apps for Apple platforms with Codex. It works over the AirDrop protocol, LAN, or Tailscale/VPN. It has some of the best performance and can go uncapped 6k@120fps no external hardware. Can even do ProRes. Streams audio too with full support for the Apple Pencil and touch with gestures. Even supports encryption and iCloud authentication so no accounts needed. It’s source available and I also open sourced the networking stack Loom. #OpenAIDevDay2026 testflight.apple.com/join/xRWsVRrv mirage.elipnik.com
English
proactive suggestions on ChatGPT Work so on point!!
English
Can anyone tell me again why would I use Fable over Kimi K3 when it's 11x cheaper?
Cline@cline
BTW how much it costs to run terminal bench with kimi k3 vs fable and gpt...
English
@punit_arani it chooses how much reasoning to use itself
i’ve had sol low take longer than sol medium if the task demands it
honestly high should work for near everything
English
I need a model reasoning effort router
I want ChatGPT, Claude and Cursor to auto-select the best reasoning effort per req
Why doesn’t this exist yet?
I always use xhigh as default and max/ultra for harder/longer tasks
I could make use of low/medium if it auto-selects it
English
@JoHei13 @mweinbach why not just use the 12 month 0% installments then resell or keep? that’s a few hundred dollars positive
English
People are going to be like "you will own nothing and be happy" but what Apple are essentially doing is giving access to hardware during the AI Memory crysis with a very good lease agreement.After 2 years if you want to buy it, it would be the same as buying it outright
Max Weinbach@mweinbach
Apple Upgrade genuinely is such a great value for some of these devices, I keep looking at devices and terms. The 16-inch MacBook Pro M5 Max with 128GB memory, 4TB storage is $250/m for 24 months. Given silicon advancements and upgrade cycles, this is a killer deal.
English