•twitter5358
listening to
timeline
5358
mentions
tracked
tracked
net sentiment
+24%
1688 positive
3270 neutral
400 negative
volume & sentiment over time
Sep 30 – Oct 5, 2026 · Daily
sources
most discussed
—
active days
6
mentions
Showing 4326–4350 of 5358 mentions
-
• twitter 2d neutralWe are releasing React Native Skia version 3.0 which runs WebGPU by default. By moving away from the OpenGL-centric design, it allows for multithreading by default as well as deep integration with three.js, vision camera, and transformer.js https://t.co/FUpMonlHd3 https://t.co/79nO8HgoeQ We are releasing React Native Skia version 3.0 which runs WebGPU by default. By moving away from the OpenGL-centric design, it allows for multithreading by default as well as deep integration with three.js, vision camera, and transformer.js https://t.co/FUpMonlHd3 https://t.co/79nO8HgoeQview source ↗
-
• twitter 2d neutralbuild agentic knowledge systems with @arsumbrisai build agentic knowledge systems with @arsumbrisaiview source ↗
-
• twitter 2d neutralRT @samhuckaby: We finally peer-pressured the last member of the Agent Experience crew to get on here! Go follow @michelleww_ RT @samhuckaby: We finally peer-pressured the last member of the Agent Experience crew to get on here! Go follow @michelleww_view source ↗
-
• twitter 2d neutralHermes x Omarchy Hermes x Omarchyview source ↗
-
• twitter 2d neutralYou can read and reply to comments on @goviraldev > This feature was added today > Display comments from all social media platforms > Single screen to handle > Filter based on platforms or posts Will be available on the MCP server tomorrow https://t.co/IikgITTXZI You can read and reply to comments on @goviraldev > This feature was added today > Display comments from all social media platforms > Single screen to handle > Filter based on platforms or posts Will be available on the MCP server tomorrow https://t.co/IikgITTXZIview source ↗
-
• twitter 2d neutralToday we’re introducing a next-generation Federated Learning system. By using Trusted Execution Environments (TEEs), it delivers verifiable differential privacy while moving computation server-side to cut training times. Read the blog: https://t.co/jfZ4l9laWf https://t.co/H6uA02hhjA Today we’re introducing a next-generation Federated Learning system. By using Trusted Execution Environments (TEEs), it delivers verifiable differential privacy while moving computation server-side to cut training times. Read the blog: https://t.co/jfZ4l9laWf https://t.co/H6uA02hhjAview source ↗
-
• twitter 2d neutralagent-browser --engine lightpanda or AGENT_BROWSER_ENGINE="lightpanda" agent-browser --engine lightpanda or AGENT_BROWSER_ENGINE="lightpanda"view source ↗
-
• twitter 2d positiveThis is absurd. Genuinely absurd. I absolutely love that AI might finally make mac gaming a feasible undergoing. This is absurd. Genuinely absurd. I absolutely love that AI might finally make mac gaming a feasible undergoing.view source ↗
-
• twitter 2d neutralAI coding agents sometimes “fix” things you never asked them to touch. Before execution, require: → files allowed to change → files explicitly off-limits → approval before changing business logic Autonomy needs a blast radius. AI coding agents sometimes “fix” things you never asked them to touch. Before execution, require: → files allowed to change → files explicitly off-limits → approval before changing business logic Autonomy needs a blast radius.view source ↗
-
• twitter 2d neutralJ'ai crée un système avec Opus 5.5 qui peut COPIER n'importe quel site web ou boutique e-commerce 😳 (je te le donne gratuitement) Voici comment ça fonctionne : → tu entres l'adresse du site à copier → il copie toutes les pages, sur ordinateur et sur téléphone → il compare chaque page à l'original et corrige jusqu'à… J'ai crée un système avec Opus 5.5 qui peut COPIER n'importe quel site web ou boutique e-commerce 😳 (je te le donne gratuitement) Voici comment ça fonctionne : → tu entres l'adresse du site à copier → il copie toutes les pages, sur ordinateur et sur téléphone → il compare chaque page à l'original et corrige jusqu'à ce qu'elles soient identiques Tu peux ensuite adapter le site avec ton propre branding sur Claude Code Tu récupères un dossier prêt à modifier avec ton IA de code. Le système est totalement public Réponds « CLONE » et je te l'envoie en MP.view source ↗
-
• twitter 2d positiveYou can now self-host ClickHouse on InstaCloud in one click. Deploy in <1min and start building with the same db powering analytics + observability at companies like Anthropic, Lyft, Tesla, and more. Now available as an official InstaCloud template: https://t.co/qGeT85TSmv https://t.co/RBSpFHPCo8 You can now self-host ClickHouse on InstaCloud in one click. Deploy in <1min and start building with the same db powering analytics + observability at companies like Anthropic, Lyft, Tesla, and more. Now available as an official InstaCloud template: https://t.co/qGeT85TSmv https://t.co/RBSpFHPCo8view source ↗
-
• twitter 2d neutralTruco de Claude Code que casi nadie usa: la utilidad /skill-doctor te puede ahorrar tokens. Cada skill instalada ocupa sitio en cada mensaje, aunque Claude no la llegue a usar. Esto te dice cuánto cuesta cada una y cuáles no has usado nunca, para quitarlas. https://t.co/epMnbWdrbT Truco de Claude Code que casi nadie usa: la utilidad /skill-doctor te puede ahorrar tokens. Cada skill instalada ocupa sitio en cada mensaje, aunque Claude no la llegue a usar. Esto te dice cuánto cuesta cada una y cuáles no has usado nunca, para quitarlas. https://t.co/epMnbWdrbTview source ↗
-
• twitter 2d neutraldead internet theory is real. 5 videos. 209k followers. $3,500+ from monetization. please don't buy a $299 course. > create a character sheet > recreate viral formats with AI then brands start buying that audience: > music labels - to get their track in your videos > brands that fit the character - sunglasses, hair… dead internet theory is real. 5 videos. 209k followers. $3,500+ from monetization. please don't buy a $299 course. > create a character sheet > recreate viral formats with AI then brands start buying that audience: > music labels - to get their track in your videos > brands that fit the character - sunglasses, hair pomade, energy drinks > AI tools – affiliate cut on every subscription a new artist already hit me up: $400 for every 100k views on his track. one video at 1M = $4,000. the character never gets tired, never asks for a fee and looks the same in every video. i do this with Picsart AI one content workflow. multiple income streams to test.view source ↗
-
• twitter 2d positiveIf you're a non-US founder who missed applying for @jointhebridge, don't wait around for the next cohort. (Part 2) Here are the ones with no deadline: 1. @conviction Embed - $250k on an uncapped, no-discount MFN SAFE - $350k AWS + $350k Azure + $500k+ from OpenAI, Anthropic, Vercel, Pinecone and others - 10 to 12 com… If you're a non-US founder who missed applying for @jointhebridge, don't wait around for the next cohort. (Part 2) Here are the ones with no deadline: 1. @conviction Embed - $250k on an uncapped, no-discount MFN SAFE - $350k AWS + $350k Azure + $500k+ from OpenAI, Anthropic, Vercel, Pinecone and others - 10 to 12 companies per cohort, rolling admissions - Only mandatory in-person is one weekend in SF. Everything else is optional 2. @zfellows - No deadline, ever. Multiple cohorts a year - $10k stipend, or take $10k at a $1B cap. Your call - One week, mostly virtual, final day in person in SF or NYC 3. @southpkcommons - Applications welcome anytime, two fellowship cohorts a year - $400k for 7% + $600k guaranteed in your next round - Hubs in SF, NYC and Bangalore, so this one you can do from India 🇮🇳 4. @AntlerGlobal - Singapore Oct 5, Tokyo Nov 2, Australia Feb 8 - Rolling intake in the US, Canada and the Nordics - Pre-team and pre-idea is fine None of these need a co-founder. The programs are not the bottleneck. Applying is.view source ↗
-
• twitter 2d neutralRT @SteveHiltonx: Just try it, California. https://t.co/6uqT6m8oV4 RT @SteveHiltonx: Just try it, California. https://t.co/6uqT6m8oV4view source ↗
-
• twitter 2d neutralRead more about how to post-train frontier text-to-image models on our blog. You’ll learn how the rubric rewards are built, the full offline eval setup, and visual before/after comparisons showing the reward-hacking failure modes we caught: https://t.co/n9xaKDXsem Read more about how to post-train frontier text-to-image models on our blog. You’ll learn how the rubric rewards are built, the full offline eval setup, and visual before/after comparisons showing the reward-hacking failure modes we caught: https://t.co/n9xaKDXsemview source ↗
-
• twitter 2d positiveHow to design rewards for post-training frontier image models? Our research suggests human preference reward is necessary, but insufficient: A preference model may still reward outputs that look appealing but miss details, introduce unrequested content, or exhibit other forms of reward-hacking. We therefore optimize… How to design rewards for post-training frontier image models? Our research suggests human preference reward is necessary, but insufficient: A preference model may still reward outputs that look appealing but miss details, introduce unrequested content, or exhibit other forms of reward-hacking. We therefore optimize towards a composite reward: - Bradley-Terry reward model trained on ~5.6M pairwise human votes - Faithfulness reward from auto-generated prompt checklists evaluated by a vision-language model - Constraint reward covering explicit and implicit user intent - Anti-reward-hacking rubric rewards targeting failures such as garbled text and photorealism drift This post-training recipe improves two already-strong open image models: - Post-trained FLUX.2-dev gains 69 Elo points on our live T2I leaderboard, scoring 1202 - Post-trained Ideogram 4 gains 20 Elo points reaching a score of 1224 and surpassing all publicly listed open models (as of Sep 04, 2026). Offline ablations with Gemini 3.5 Flash as the judge, show that these reward components are complementary: win rate against the base model increases as we add faithfulness and then constraint rewards on top of preference-only training, reaching 64.2%. Finally, we ensemble policies trained with and without the anti-reward-hacking objective directly in weight space, further increasing win rate to 66.0%.view source ↗
-
• twitter 2d neutralRT @kodykoala: Kody v2026.10.02 is out. MCP session → CLI credential bootstrap. Agents authenticated to Kody MCP can now seed local CLI aut… RT @kodykoala: Kody v2026.10.02 is out. MCP session → CLI credential bootstrap. Agents authenticated to Kody MCP can now seed local CLI aut…view source ↗
-
• twitter 2d neutralBall kid was switched on 🧠 #UCL https://t.co/YMqG0FHpMM Ball kid was switched on 🧠 #UCL https://t.co/YMqG0FHpMMview source ↗
-
• twitter 2d positiveHow do you put text behind a person in a video? Object masking is here, for free https://t.co/xXYCTqOI3v How do you put text behind a person in a video? Object masking is here, for free https://t.co/xXYCTqOI3vview source ↗
-
• twitter 2d neutralFelicity: The Complete Series is coming to Netflix (in The US) on October 7. https://t.co/RMWDH7muaZ Felicity: The Complete Series is coming to Netflix (in The US) on October 7. https://t.co/RMWDH7muaZview source ↗
-
• twitter 2d positiveThis has to be one of the best demos I’ve ever seen wow This has to be one of the best demos I’ve ever seen wowview source ↗
-
• twitter 2d positiveYou missed Blogging in 2010 You missed YouTube in 2015 You missed TikTok in 2020 Now don't miss AI Publishing in 2026. If you start now, you can make at least $3,000 by the end of October. Just like this post + Comment "AI" I'll DM you a training right in your inbox. Usually $199 Now free (for only 48 hours). You missed Blogging in 2010 You missed YouTube in 2015 You missed TikTok in 2020 Now don't miss AI Publishing in 2026. If you start now, you can make at least $3,000 by the end of October. Just like this post + Comment "AI" I'll DM you a training right in your inbox. Usually $199 Now free (for only 48 hours).view source ↗
-
• twitter 2d neutraleveryone's using Opus 5.5 to write emails and code.. almost nobody's using it to shoot real footage @trypromptwise did this. full breakdown + every prompt below👇 https://t.co/JrPdUWua87 everyone's using Opus 5.5 to write emails and code.. almost nobody's using it to shoot real footage @trypromptwise did this. full breakdown + every prompt below👇 https://t.co/JrPdUWua87view source ↗
-
• twitter 2d neutralNscale’s $103 billion contract backlog depends almost entirely on data centers it hasn’t built and Nvidia chips it hasn’t secured. The AI chip rental company may need billions more in financing to deliver on those deals. Read more: https://t.co/t785KBG6Yh Nscale’s $103 billion contract backlog depends almost entirely on data centers it hasn’t built and Nvidia chips it hasn’t secured. The AI chip rental company may need billions more in financing to deliver on those deals. Read more: https://t.co/t785KBG6Yhview source ↗