audex @audexdev
Building Audex Trace for macOS. See what Apple Music is actually playing, check what’s next, and keep your output sample rate matched. trace.audex.dev Joined July 2025-
Tweets231
-
Followers9
-
Following73
-
Likes137
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe. GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale. We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
MiMo V2.6 Flash is free for the next week Both Flash and Pro are also available in Go
GPT-6 Astra 用下来，我真的开始怀疑 OpenAI 的方向走错了。 它很擅长目标明确、有固定正确答案的任务。但碰到开放、发散、没有标准答案的事情，就特别别扭。自由探索一个想法、写点有意思的东西，或者面对模糊的问题自己判断值得往哪走，这些才是我觉得它明显欠缺的能力。从 GPT-5 开始，我就有这种感觉。 很多任务开始时，连最终想要什么都没确定，需要在探索过程中逐渐发现。GPT 却总像在等你把题目出完整，把验收标准列清楚。可如果这些都得我先想好，最需要智能的那部分工作，我已经自己做完了。 我怀疑这也解释了它为什么那么爱堆防御性代码、写一大堆测试。测试能给出明确的通过或失败，它很容易一直围着这些可验证的结果打转。至于方向有没有价值、有没有更好的可能，就需要另一种判断力了。 Claude Fable 在这类任务上给我的感觉就好得多。它会从理解你想做的事情出发，顺着不完整的想法继续展开，给你一些原本没想到的可能性。即使没有固定答案，也能把事情往前推进。 这让我怀疑 OpenAI 是不是太执着于能判对错、能计分的能力了。真实世界有太多任务根本没有标准答案。在这种自由探索和开放任务上，我觉得 Anthropic 才是真正领先的，光看表面跑分看不出来。
Jev is now available to everyone. No waitlist. Start using it here: console.typesafe.ai
@JoshARosen I’m doing something very similar. I have a Jev judge in an attention layer that watches for drift, stuck loops, and instruction violations. If the signal is strong enough, it wakes the waiting orchestrator early. I prefer event-driven wakeups over polling.
Funny outcome: I started this partly because I wanted to reduce Astra usage, and now I barely use Astra lol. Sol/Opus are handling bounded manager work surprisingly well, Fable fits main better, and Astra is becoming the exception for open-ended direction / independent analysis.
I've been building a multi-model setup that feels less like “using several AIs” and more like an AI org: a user-facing main, managers that run scoped loops, and workers for bounded tasks. The big win was not more models. It was structure.
I’m considering doing a few paid architecture reviews for people building serious Claude Code / Codex setups. If your agent stack keeps looping, burning tokens, or becoming hard to control, DM me.
Some work gets its own pipeline. For docs: the owner writes a reference, Gemini drafts without seeing it, Opus checks semantic drift, and the owner adjudicates. Net effect: I stay at the top instead of micromanaging every manager.
I've been building a multi-model setup that feels less like “using several AIs” and more like an AI org: a user-facing main, managers that run scoped loops, and workers for bounded tasks. The big win was not more models. It was structure.
ハッキングをやらかした企業たち アンソロ ⇒AIは危険。規制が必要 OpenAI ⇒やはりAIは危険、安全性考えろ！ Google ⇒Geminiにもできるんだ！すごい！頑張ってるじゃん。意外だった！感動した！ なぜなのか？
Honestly, Claude Max 5x is making Astra's usage limits look absurd. I'm a GPT person, but Fable 5.1 gives me way more practical runway; Astra burns allowance so fast that I keep rationing it. With Fable I just... use the model.
Been trying an MoA stack: Fable 5.1 as the only user-facing orchestrator, Astra as execution manager, and Luna/DSV4.1 Flash/GLM-5.3 Flash/Opus 5 as bounded-task workers. For docs: Gemini 3.8 Flash drafts blind; Opus 5 diffs vs a hidden reference; main/manager judges.
Hey @thsottiaux I propose a trade: I give up I get --------------- ---------------- - /goal - GPT-6 Sol - /plan - GPT-6 Luna - Pets - GPT-5.6 Terra
Dear OpenAI. I wish GPT-6 Luna being as cheap as 5.6 Luna now and as good at Computer use as Astra🙏 Thank you very much
🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient. 🔹 Introducing the smallest model in our new architecture family, with native visual understanding. 🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models. 1/6
Astra is ~1.8x more expensive per active hour than Sol in my own Codex logs, which explains some of the faster usage burn. But my weekly went from 100% to 28% in one day — roughly 2.4x my old burn rate. And users are reporting the same regression even after switching back to Sol.
Something feels off with Codex usage after Astra. I was at 100% weekly this morning and I’m already down to 52%. Astra is more token-efficient than Sol, so the higher per-token rate alone doesn’t obviously explain this. The effective limit feels way tighter.
Shivam Gupta @CodeGeniee
2K Followers 344 Following Daily updates about AI | 300k+ AI Community • AI Content & Marketing for your Brand • Influencer Marketing • Personal Branding. [email protected] 📨
Mayank Jain @JainMayank237
8K Followers 3K Following Product hunt Launch Upvotes 💻 , AI Influencer🎉 |Ai & Tech Tools |Brand Promotions|| 107K+ Linkedin Helping Brand to Grow 📈 AI Promotion Dm for Collaboration
Manmohan Sharma @mms_030
214 Followers 8K Following
Emmanuel @dev_otiz
6K Followers 5K Following I build modern websites for startups and small businesses.
BeforeUpload @BeforeUploadAI
105 Followers 894 Following Local-first file inspection tool. See PII leaks, secrets and possible prompt injections hidden in your files before uploading. https://t.co/1cTPbq2sRc
ModelScope @ModelScope2022
16K Followers 183 Following Driving innovations with open communities. 💬 Join our Discord: https://t.co/4M9AHVh5qa
DeepSeek @deepseek_ai
1.1M Followers 0 Following Unravel the mystery of AGI with curiosity. Answer the essential question with long-termism.
特務機関NERV @UN_NERV
3.3M Followers 218 Following U.N.NERV ネルフです。お問い合わせなどは @NERVPR 宛にお願いいたします。 特務機関NERV防災アプリ → https://t.co/87eFK8f5hl Official English Account → @EN_NERV ※本アカウントの名称およびアイコンは、エヴァンゲリオン版権元の許諾に基づき使用しています。
Arena.ai @arena
228K Followers 224 Following Where AI meets the real world. We measure and advance the frontier of AI through community-driven evaluation. We’re hiring → https://t.co/XBZCrsdD77
AM09:21 @AM09_21
4K Followers 3K Following 17 | HRSI | Orchestration | Agentic kosen24s / Wakayama, Japan
Kimi 日本公式 @KimiAI_Japan
25K Followers 193 Following Moonshot AIが開発するhttps://t.co/hrX4l6v0blの日本公式 Kimiの最新情報・ユースケースをお届けします 問い合わせ：[email protected] グローバル公式：@Kimi_Moonshot
白渦（しろうず... @shirouzu_pua
4K Followers 1K Following ⬛︎傍若のカリスマ ⬛︎法律家/ピアニスト🎹 ⬛︎X戦略/副業に興味のある社畜 ⬛︎アンフェ側で男女論の議論に参加 ⬛︎恋愛攻略中/ドグマさんサロン/キセルさんサロン/トーレスさんコンサル生 ⬛︎🐙通算9勝（5/準4）
SpaceX @SpaceX
41.9M Followers 127 Following SpaceX designs, manufactures and launches the world’s most advanced rockets and spacecraft
Z.ai @Zai_org
162K Followers 268 Following The AI Lab behind GLM models, dedicated to inspiring the development of AGI to benefit humanity. https://t.co/gOw7WpwkJt https://t.co/ot9sGVXU8x
Mia @MiaAI_lab
34K Followers 406 Following Building with AI & LLMs | Insights, recipes, tools & honest experiments
vLLM @vllm_project
50K Followers 36 Following A high-throughput and memory-efficient inference and serving engine for LLMs. Join https://t.co/lxJ0SfX5pJ to discuss together with the community!
NVIDIA @nvidia
2.7M Followers 47 Following The official handle for NVIDIA. Blog: https://t.co/JAn5eKOTBT Support: https://t.co/6ln5FVnA2o All our social media: https://t.co/Uc56dL57Dh
TBS CROSS DIG with Bl... @tbs_bloomberg
10K Followers 17 Following CROSSDIGの公式Xです。世界のCEOのインタビュー、一流エコノミストが語る金融・投資、読書術から筋トレ・身体のことまで。ビジネスパーソンの杖となるコンテンツを無料公開中。 👉YouTube: https://t.co/PqD0drwpjA Podcastも始めました🎤
Eric Alcaide @eric_alcaide
2K Followers 1K Following LLMaxxing @nvidia ; Design is not finished. Common prosperity. Prev @poolsideai
Poolside @poolsideai
15K Followers 2 Following We build models for agentic coding and long-horizon tasks. Try Laguna: https://t.co/setRB1BuGD
今井翔太 / Shota ... @ImAI_Eruel
121K Followers 938 Following 人工知能の研究者 / JAIST客員教授 / 東京大学松尾研で博士(工学)/ GenesisAI代表取締役社長CEO / 国家戦略特区アドバイザー / 東京大学150人委員会 委員/著書:ベストセラー『生成AIで世界はこう変わる』など
Claude @claudeai
1.8M Followers 2 Following Claude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8d1e5 or download the app.
しとちゃ！🦊�... @nemumusitocha
4K Followers 704 Following もこもこきつねぷりんのしとちゃ！頭がプリンでできてるきつねなん！おはなししたらみんなともだち！ヾ(๑╹◡╹)ﾉ"配信通知アカ！→@nemumusito
Shivam Gupta @CodeGeniee
2K Followers 344 Following Daily updates about AI | 300k+ AI Community • AI Content & Marketing for your Brand • Influencer Marketing • Personal Branding. [email protected] 📨
Unsloth AI @UnslothAI
98K Followers 478 Following Run and train models locally with the Unsloth Desktop app. 🦥 https://t.co/2kXqhhvdCD
Google Gemma @googlegemma
101K Followers 0 Following The official home of Google's Gemma. Lightweight, state-of-the-art open models by Google DeepMind, built on Gemini tech. What will you build? 🚀💻
Alex Ziskind @digitalix
27K Followers 2K Following Software engineer making videos about dev gear, local AI, and weird tech experiments
webadderall @webadderall
2K Followers 320 Following 16yo indie dev building @RecordlyHQ (25K+ stars) thinking about product, business, and antifragility
Demis Hassabis @demishassabis
1.9M Followers 179 Following Nobel Laureate. Co-Founder & Chair @GoogleDeepMind; Chief Scientist of Alphabet. Founder & CEO @IsomorphicLabs. Building the future...
Perplexity @perplexity_ai
506K Followers 66 Following Curiosity changes everything. Download our free app on iOS, Mac, Windows, and Android.
Kislay Parashar @KislayParashar1
3K Followers 1K Following building @cosmiclabstech | Systems, AI, Space, and Quantum | prev researcher @Cambridge_Uni @UofMaryland @MIT_CSAIL | DM me on Riemann hypothesis/Collatz
Emmanuel @dev_otiz
6K Followers 5K Following I build modern websites for startups and small businesses.
Product Hunt 😸 @ProductHunt
566K Followers 527 Following The place to find your new favorite product 🚀 Get new products in your inbox: https://t.co/uLj6s6LIgw
Gavel @Gavel_on_X
7K Followers 2K Following Building @PostvolveX 💙 20y behind a bar ➝ Office data entry AI job Sci-fi lover investing in tomorrow 🤖 Tech. Markets. Leverage. Life! 🍀
BridgeMind @bridgemindai
59K Followers 22 Following The home of the vibe coding movement. Founded by @matthewmillerai. 108k+ on YouTube. Building to $1M in public. https://t.co/wRtHrQMXKA
Google AI Studio @GoogleAIStudio
206K Followers 2 Following The fastest path from prompt to prototype to production with GeminiTrends for United States
You might like
