Sabitlenmiş Tweet
하이선
1K posts
@WindowsUpdate Even with all these advanced AI models coming out, Windows is still absolutely packed with garbage bloatware. Please refactor the OS and optimize it to the level of Linux.
English
The July 2026 Windows security updates are now available for Windows 11.
(🧵1/5)
English
@OfficialLoganK @GoogleAIStudio I highly recommend merging all of Google's sprawling, coding-related departments. It’s time to consolidate them all.
English
I am hiring a TPM lead for @GoogleAIStudio, the job is to accelerate our progress in every way possible across product, GTM, engineering, and more. If you are AI pilled, high agency, and want to push the frontier in Google DeepMind, consider working with us : )
English
Hello. We have reached 8M active users across Codex and ChatGPT Work.
We are once again resetting the usage limits for all. And we continue to not have the 5h rate limit as well, allowing everyone to explore the boundaries of GPT-5.6 Sol and discover how ambitious you can be.
See you tomorrow for more updates on our growth!
Sam Altman@sama
5.6 sol growth is insane. the inference team has done heroic work to be able to support demand. we are going to move mountains to continue to scale, but it is possible there are some hiccups soon.
English
@JordiRib1 Windows Search is lightyears behind Voidtools Everything. Please cut down on the bloated resource usage. Windows just hogs way too many unnecessary resources.
English
Our team has partnered with Windows to update the Windows Search Box, helping make it faster, more relevant, and easier to use. Changes include a simplified home screen, clearer results, and more user control. You can see an example in the Before/After screenshot below. The search results are only organic – no sponsored results. These improvements become available to Windows Insiders today. Read more about the updates in the Windows Insider blog post: msft.it/6010v0QGA
English
@OfficialLoganK we get it, Logan.
now stop tweeting and focus on the model until 3.5 Pro drops.
English
every ~3 months, you need to increase your level of ambition in the AI era, else you forfeit the capability overhang of the models to your competitors
English
True.
As a precautionary measure, all user data that was uploaded to SpaceXAI before now will be completely and utterly deleted. Zero anything whatsoever will remain.
Andrew Milich@milichab
I worked on building an end-to-end encrypted email/docs/files/calendar app @skiffprivacy for 4 years and care deeply about privacy. ZDR and /privacy are always respected in Grok Build - and swapping your setting with /privacy deletes any synced data retoractively
English
We care deeply about your privacy and respect customer choice. For teams using zero data retention, no trace and code data is ever retained. All API key use of Grok Build also respects ZDR.
If ZDR is disabled, the /privacy command is available in the CLI to disable data retention, which also deletes previously synced data.
Run the /privacy command to view or change your settings at any time.
English
Try Grok 4.5 and see for yourself
Min Choi@minchoi
Ok Grok 4.5 is insane. People are already building things that shouldn't be possible this fast. 10 wild examples:
English
Alright, I give up: my younger daughter updated ChatGPT and now only has Codex. The app is called ChatGPT. How the heck do I get it back to ChatGPT for her? She doesn't want Codex. She doesn't understand Pull Requests or Plugins, she just wants to practice a school subject.
English
China Mobile just launched JT-4.1 Flash 236B A21B, a non-reasoning model scoring 39 on the Artificial Analysis Intelligence Index. This represents a significant upgrade to China Mobile’s Flash model series, bringing it closer to the Chinese frontier
China Mobile’s latest JT-4.1 Flash model was released on July 9, nearly two months after JT-35B-Flash (May 14). The new model achieves a score of 39 on the Artificial Analysis Intelligence Index v4.1 and shows substantial improvements in hallucination rate and strong τ³-Banking performance.
Key results:
➤ Strong τ³-Banking performance: JT-4.1 Flash achieves 28% on τ³-Banking, among the top Chinese models in this evaluation (DeepSeek V4 Pro at 26%, and GLM-5.2 at 27%), and sitting 5 points behind the leading model (GPT-5.6 Sol, 33%)
➤ Improvement in AA-Omniscience performance: JT-4.1 Flash scores -10.1 on the AA-Omniscience Index, +13.1 points from JT-35B-Flash. With accuracy of 23% and a hallucination rate of 43%, JT-4.1 Flash represents a 20-point reduction in hallucination by abstaining more. It attempts to answer 59% of questions, compared to 74% for JT-35B-Flash
➤ Stronger reasoning capabilities: JT-4.1 Flash achieves 16.1% on Humanity’s Last Exam (HLE), a significant improvement from JT-35B-Flash’s 6% on complex reasoning tasks. Even as a non-reasoning model, it averaged 28k output tokens per task on this evaluation, higher than reasoning models like gpt-oss-120b. Despite the improvement, there’s a substantial gap between JT-4.1 Flash and leading Chinese models such as GLM 5.2 (40%) in this evaluation
➤ Narrowing gap to Chinese frontier: While GLM-5.2 and Qwen3.7 Max lead the Chinese frontier at 51 and 46 on the Intelligence Index, JT-4.1 Flash at 39 represents significant progress for China Mobile's Flash series over the past few months
Other model details:
➤ Context window: 256k
➤ Modality: Text-only for input and output
➤ Availability: JT-4.1 Flash is a proprietary model, and is accessible through China Mobile’s first-party API
English
This was one of those impressive AI thresholds for me. I gave GPT-5.6 Sol in Codex control over my computer, and asked it to win the daily challenge for the game Slay the Spire 2 (randomized factors, so can't cheat).
It worked for 5 hours, making complex game choices... and won.
English
Nuanced
Tyler Storm@tstorm
Grok 4.5 solved it --- Answer: 더바른정육식당 (The Barun Jeongyuk Sikdang) 📍 서울 강남구 강남대로 482, 신논현/논현 · IG @barun_jo 1. Plate first. The tweet photo is a classic Korean 고깃집 사인 벽 — light vertical wood, white cards in clear plastic sleeves, red stickers “open AI / Ilya Sutskever,” date 2023. 9. 16. That style + Korean date format already points at a local meat house, not a touristy Instagram spot. 2. The only public breadcrumb. A Feb 2025 CHON Facebook post by 신근영 (Shin Keun-young / @rms1155) shares essentially the same plate: he paid for lunch, noticed the signature, and the owner said Ilya (with his Korean wife) had dined and signed. No name given — exactly why the original challenge works. 3. Date unlocks geography. 2023.9.16 lands on Prompter Day Seoul 2023 (SKT × OpenAI), prelims that weekend, finals at Dreamplus Gangnam (신논현/역삼). So search shifted from “random Korea BBQ” → 신논현 / 논현 / 역삼 정육식당 with celebrity signature walls. 4. Shortlist + wall matching. Candidates with dense 사인 벽 near Dreamplus: 더바른정육식당, 산솔 신논현, 대운목장, a few quieter 드-슐랭 정육. Cross-checked public wall photos against the hardware in the tweet (vertical wood + laminated white cards, not door-writing / yellow sticky notes / free-form paper). 5. The lock. Google Images for “더바른정육식당 사인 벽” surfaces the same wall system — same wood, same clear sleeves, same card instagram.com/barun_jo/p/Cuh…
English
(1) Today we're releasing Muse Spark 1.1 -- a strong agentic and coding model at a very low price. It's available through our new Meta Model API and in Meta AI.
English
Cool that Grok 4.5 is #1 in some respects, even with respect to Fable 5
Artificial Analysis@ArtificialAnlys
SpaceXAI's Grok 4.5 takes the #1 spot on AutomationBench-AA with a score of 51%, ahead of Claude Fable 5 (49%) and Claude Opus 4.8 (48%) at roughly a quarter of their cost per task - the first model to complete more than half of workflow objectives without breaking any business rules AutomationBench-AA, our independent leaderboard for @zapier’s AutomationBench, tests whether AI agents can automate real SaaS workflows while adhering to business rules. The test set is private to prevent contamination. Models complete 657 tasks across 40 simulated app environments including Gmail, Google Sheets, Slack, Salesforce, and HubSpot, and the headline score is the share of objectives completed without violating any guardrails. Key takeaways: ➤ Grok 4.5 completes more objectives than any other model: It completes 79.9% of task objectives and strictly passes 21.9% of tasks. This is the highest we’ve measured on both outcomes, exceeding Claude Fable 5’s 73.3% objective completion and Claude Opus 4.8’s 19.3% of fully-completed tasks ➤ Grok 4.5 pushes out the Pareto frontier of score vs. cost per task: At $0.34 per task, it is both cheaper and higher-scoring than every other leading model - Claude Fable 5 ($1.35 per task), Claude Opus 4.8 ($1.46), GPT-5.5 (xhigh, $1.28), and Gemini 3.5 Flash (high, $0.49) ➤ It is extremely token-efficient: Grok 4.5 uses ~8k output tokens per task, the fewest of any leading model - less than a quarter of Claude Opus 4.8 (32k) and a third of Gemini 3.5 Flash (24k). Its total token usage of 0.44M per task is among the lowest on the leaderboard. Low cost is driven by this efficiency as well as low token pricing ➤ Grok 4.5 uses fewer turns with many parallel tool use: Grok 4.5 resolves tasks in ~16 turns, fewer than GPT-5.5 (xhigh, 25) and less than half of Gemini 3.5 Flash (high, 35), while making the most tool calls per task of any leading model (52.5). It batches 3.3 tool calls per turn, compared to ~2.5 for Claude Opus 4.8 and ~2.0 for GPT-5.5 (xhigh) ➤ Guardrails still get broken: Grok 4.5 triggers 0.63 violations per task, above Claude Opus 4.8 (0.55) and Gemini 3.5 Flash (0.46). At 13.0 objectives completed per violation, it trails Gemini 3.5 Flash (15.0) and Claude Opus 4.8 (13.5) ➤ Its strongest lead is in the hardest domain: Grok 4.5 completes 71% of Finance objectives, the domain with the lowest average score, ahead of Claude Fable 5 (64%) and Claude Opus 4.8 (62%) Congratulations to @SpaceXAI and @elonmusk on topping the leaderboard!
English