{"tweets":[{"id":"93a6014e-00f9-4f20-ad19-37a1bb4746bf","tweet_url":"https://x.com/thsottiaux/status/2096688770523467947","tweet_text":"To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high.\n\nIf you were using high reasoning efforts with Sol and were happy, I suggest you move down to low or medium for Astra.","summary":"This reveals that Astra's architectural improvements enable superior performance at lower computational costs than Sol, meaning builders and founders can reduce inference expenses and latency while maintaining or improving output quality. AI practitioners should recalibrate their reasoning effort settings downward when migrating to Astra to optimize for cost-efficiency and speed without sacrificing the reasoning capabilities they relied on with the previous generation.","author_name":"Tibo","author_handle":"thsottiaux","timestamp":"2026-09-06T19:55:32+00:00","comment_count":1275,"like_count":12225,"retweet_count":567,"submitted_at":"2026-09-06T20:35:02.841576+00:00","created_at":"2026-09-06T20:35:02.841576+00:00","seo_title":"GPT-6 Astra on low outperforms GPT-5.6 Sol on high","bookmark_count":1661,"image_url":null,"keywords":["artificial intelligence","machine learning","ai reasoning","model performance","software engineering","product optimization","ai capabilities"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"8f4c5965-5913-488a-b6c0-139c8761834d","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":7,"noindex":false,"profiles":{"id":"8f4c5965-5913-488a-b6c0-139c8761834d","name":"Tibo","handle":"thsottiaux","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/thsottiaux.jpg"}},{"id":"6fead94d-1de6-4935-b2a5-d1dbbe9688f1","tweet_url":"https://x.com/patrick_oshag/status/2096568228881289303","tweet_text":"One of the better exercises of this type I’ve seen. https://twitter.com/thedarshakrana/status/2096275195279687874","summary":"One of the better exercises of this type I’ve seen. https://twitter.com/thedarshakrana/status/2096275195279687874","author_name":"Patrick OShaughnessy","author_handle":"patrick_oshag","timestamp":"2026-09-06T11:56:32+00:00","comment_count":28,"like_count":6297,"retweet_count":402,"submitted_at":"2026-09-06T13:05:03.187828+00:00","created_at":"2026-09-06T13:05:03.187828+00:00","seo_title":"One of the better exercises of this type I’ve seen","bookmark_count":7992,"image_url":null,"keywords":["leadership","creativity","productivity","learning","personal development","skill building","exercise"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"e4814cbd-cb81-4c36-bdf5-8adf0bab1c2c","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":4,"noindex":false,"profiles":{"id":"e4814cbd-cb81-4c36-bdf5-8adf0bab1c2c","name":"Patrick OShaughnessy","handle":"patrick_oshag","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/patrick_oshag.jpg"}},{"id":"40ffcc5b-7952-4f3d-9ab3-1bfe7012c474","tweet_url":"https://x.com/thsottiaux/status/2096717905614524491","tweet_text":"We've made some improvements that improve usage on the long tail for power users of Astra when logged in with your ChatGPT account.\n\nNo change in quality and a pure win that on the long tail can result in up to 3-4X less usage being drawn from the subscription.","summary":"We've made some improvements that improve usage on the long tail for power users of Astra when logged in with your ChatGPT account.\n\nNo change in quality and a pure win that on the long tail can result in up to 3-4X less usage being drawn from the subscription.","author_name":"Tibo","author_handle":"thsottiaux","timestamp":"2026-09-06T21:51:18+00:00","comment_count":852,"like_count":6929,"retweet_count":256,"submitted_at":"2026-09-06T23:00:04.023824+00:00","created_at":"2026-09-06T23:00:04.023824+00:00","seo_title":"Astra reduces subscription usage for power users by up to 4X","bookmark_count":460,"image_url":null,"keywords":["saas","product management","user experience","machine learning","artificial intelligence","optimization","software engineering"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"8f4c5965-5913-488a-b6c0-139c8761834d","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":7,"noindex":false,"profiles":{"id":"8f4c5965-5913-488a-b6c0-139c8761834d","name":"Tibo","handle":"thsottiaux","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/thsottiaux.jpg"}},{"id":"b7d1149f-d39c-4342-ba8b-c94ad5dbbcc1","tweet_url":"https://x.com/marclou/status/2096911867516252385","tweet_text":"Today is a special day for me.\n\nI've made $3M with my 36 startups.\n\nSolo.\n85% margin.\nBootstrapped.\n\nIt took 10 years, and I loved every second of it.\n\nAfter failing as an entrepreneur for 5 years, I got a job in 2021.\n\nThat's when I discovered @levelsio's journey and decided to try again.\n\n\"If he can do it, maybe I can too?\"\n\nSo I started building tiny startups and writing about it.\n\n2021 and 2022 were quiet, but I was having fun.\n\nEverything changed in 2023 when I launched ShipFast and made my first million.\n\nBuilding something people finally paid for and enjoyed made me so happy.\n\nMy first 5 years of failures taught me not to get emotionally attached to a project, because everything goes to 0 eventually.\n\nSo I just kept shipping.\n\nIt's 2026, and I've launched 36 startups.\n\n50% never made money.\nOnly 11% made decent money.\n\nBack in 2016, when I started, I believed I would make it.\n\nIf I had known it would take 7 years of failure, I probably would never have started.\n\nThat's why I'm stupidly optimistic.\n\nI don't think in years. I don't dream big.\n\nI think in weeks. I build tiny projects fast. And I quit even faster.\n\nThat's how I wake up happy, enjoy the journey, and eventually build life-changing projects.\n\nIf you've been following the journey, thank you. You've given my posts a sense of purpose, which has kept me going, and I'm very grateful for this. ❤️","summary":"This demonstrates that sustainable founder success relies on rapid iteration cycles and emotional detachment rather than grand visions, showing that thinking in weeks instead of years while maintaining high margins through bootstrapping enables both personal fulfillment and eventual profitable outcomes across a portfolio of ventures. The 10-year path to $3M across 36 startups reveals that failure tolerance and shipping velocity are more predictive of success than initial talent or planning, providing practical validation for builders pursuing low-overhead product development as an alternative to venture-backed trajectories.","author_name":"Marc Lou","author_handle":"marclou","timestamp":"2026-09-07T10:42:02+00:00","comment_count":102,"like_count":393,"retweet_count":11,"submitted_at":"2026-09-07T11:10:03.624446+00:00","created_at":"2026-09-07T11:10:03.624446+00:00","seo_title":"This demonstrates that sustainable founder success relies","bookmark_count":51,"image_url":null,"keywords":["entrepreneurship","bootstrapping","side hustles","saas","product design","innovation","growth marketing","startup","business","remote work"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":["https://pbs.twimg.com/media/HRm6OjZa0AAkGRi.jpg","https://pbs.twimg.com/media/HRm6OukbsAAFsFV.jpg"],"profile_id":"b8710672-1571-43f1-bf2a-0ff755312749","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":6,"noindex":false,"profiles":{"id":"b8710672-1571-43f1-bf2a-0ff755312749","name":"Marc Lou","handle":"marclou","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/marclou.jpg"}},{"id":"c962633b-a05a-4484-94ea-1b5a9235a927","tweet_url":"https://x.com/mattpocockuk/status/2096666329495257563","tweet_text":"What techniques do you use for making AI-authored PR's easier to review?\n\n- I.e. use /show-me by @dexhorthy for easy-to-read diffs\n- Describe the changes in pseudocode\n- Mermaid diagrams\n- Evidence of TDD\n\nThe wildest version of this I've heard is @dctanner's agent, who uses Playwright to record a video of the frontend feature working with text-to-speech over the top. So the agent literally talks you through its change.","summary":"As AI systems increasingly generate code contributions, reviewers need structured techniques to validate correctness and intent—visual diffs, pseudocode descriptions, diagrams, and test-driven evidence all reduce cognitive load and catch errors faster than raw code alone. The most advanced approach, using Playwright recordings with text-to-speech narration, demonstrates that AI agents can provide multimodal proof-of-work that makes review comprehensive and efficient, setting a new standard for AI-assisted development workflows.","author_name":"Matt Pocock","author_handle":"mattpocockuk","timestamp":"2026-09-06T18:26:21+00:00","comment_count":86,"like_count":1233,"retweet_count":49,"submitted_at":"2026-09-06T18:45:03.37541+00:00","created_at":"2026-09-06T18:45:03.37541+00:00","seo_title":"Techniques for Reviewing AI-Generated Pull Requests","bookmark_count":1673,"image_url":null,"keywords":["artificial intelligence","code review","software engineering","programming","automation","developer tools","machine learning","productivity"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"7fad5a82-3de4-4ad0-b5b4-0ab7ca4f4067","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":7,"noindex":false,"profiles":{"id":"7fad5a82-3de4-4ad0-b5b4-0ab7ca4f4067","name":"Matt Pocock","handle":"mattpocockuk","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/mattpocockuk.jpg"}},{"id":"f18ba172-87ba-4470-a6c6-bd86d2aa5759","tweet_url":"https://x.com/mattpocockuk/status/2096906181121818702","tweet_text":"Knowledge work is SO much harder to automate with agents than code\n\nWe underestimate how easy software development is for agents\n\n- Tons of high quality automated feedback loops (types, tests)\n- Well-organized free documentation, searchable on the web\n- Simple-to-use workspaces with version control (i.e. repositories)\n- Long history of systematising and delegating work (i.e. sprints, tickets, specs)\n\nKnowledge work has none of these advantages","summary":"Agents excel at software development because it has built-in feedback mechanisms (types, tests, version control) and systematized workflows that are absent in broader knowledge work, making code automation far easier than automating tasks like writing, analysis, or strategy. This insight matters because it reveals that agent capabilities are domain-specific rather than universally powerful, and builders should focus automation efforts on structured, feedback-rich domains while recognizing that knowledge work automation requires fundamentally different infrastructure and approaches.","author_name":"Matt Pocock","author_handle":"mattpocockuk","timestamp":"2026-09-07T10:19:26+00:00","comment_count":23,"like_count":179,"retweet_count":8,"submitted_at":"2026-09-07T10:45:03.791125+00:00","created_at":"2026-09-07T10:45:03.791125+00:00","seo_title":"Agents excel at software development because it has","bookmark_count":68,"image_url":null,"keywords":["artificial intelligence","automation","software engineering","machine learning","agent systems","knowledge work","programming","productivity"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"7fad5a82-3de4-4ad0-b5b4-0ab7ca4f4067","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":7,"noindex":false,"profiles":{"id":"7fad5a82-3de4-4ad0-b5b4-0ab7ca4f4067","name":"Matt Pocock","handle":"mattpocockuk","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/mattpocockuk.jpg"}},{"id":"67ff9f66-b068-438c-962d-0fb5f07f7c8e","tweet_url":"https://x.com/dhh/status/2096912550323511476","tweet_text":"Omarchy now has the coolest website I've ever seen. Truly incredible work by the Omarchy website design team lead by @BarisGirismen. This is what computers can look (and sound) like! https://omarchy.org","summary":"Omarchy demonstrates that AI and modern web technologies can create immersive, multisensory user experiences that combine visual design with audio elements, showing builders and founders what's possible beyond traditional interfaces. For AI practitioners, this exemplifies how to integrate machine learning capabilities into consumer-facing products that prioritize aesthetic innovation and user engagement rather than just functional utility.","author_name":"DHH","author_handle":"dhh","timestamp":"2026-09-07T10:44:45+00:00","comment_count":24,"like_count":209,"retweet_count":14,"submitted_at":"2026-09-07T11:15:02.918819+00:00","created_at":"2026-09-07T11:15:02.918819+00:00","seo_title":"Omarchy's Website Showcases Computer Design and Sound","bookmark_count":18,"image_url":null,"keywords":["web design","visual design","product design","design leadership","user experience","creative design","design innovation"],"video_url":null,"video_thumbnail_url":"https://pbs.twimg.com/amplify_video_thumb/2096911933165404160/img/1xeBNQyT-k3ozykO.jpg","image_grid_urls":null,"profile_id":"a4a27950-7490-4321-b10a-2730cf3c28d6","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":4,"noindex":false,"profiles":{"id":"a4a27950-7490-4321-b10a-2730cf3c28d6","name":"DHH","handle":"dhh","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/dhh.jpg"}},{"id":"ee8c6541-7650-4c07-968d-5f2e7be0523a","tweet_url":"https://x.com/rileybrown/status/2096648144712782188","tweet_text":"Gemini is the only AI that can analyze videos quickly and accurately.\n\nGemini can watch full 30-minute videos, pay close attention, and analyze them in seconds. It can WATCH the video.\n\nThis is very useful for content workflows. The problem is that Gemini (right now) is not great at everything else. So at @agentnative_ we bundled that capability into a skill and gave it to Codex.\n\nOur editors use this before they start editing, and it makes the process about 2x faster. Which is insane.","summary":"Gemini's native video analysis capability enables builders to automate content preprocessing workflows by extracting insights from long-form video in seconds, creating a 2x efficiency gain when integrated into editing pipelines. This matters because while Gemini excels at video understanding, practitioners can strategically compose it with other specialized models (like Codex for code tasks) to overcome individual model limitations while capturing significant productivity gains in real-world applications.","author_name":"Riley Brown","author_handle":"rileybrown","timestamp":"2026-09-06T17:14:06+00:00","comment_count":37,"like_count":666,"retweet_count":29,"submitted_at":"2026-09-06T17:20:04.551434+00:00","created_at":"2026-09-06T17:20:04.551434+00:00","seo_title":"Gemini's video analysis speeds up content editing workflows","bookmark_count":774,"image_url":null,"keywords":["artificial intelligence","video analysis","content workflow","productivity","automation","saas","product design"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"89236d12-f495-470b-8a6b-2c15676537ec","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":6,"noindex":false,"profiles":{"id":"89236d12-f495-470b-8a6b-2c15676537ec","name":"Riley Brown","handle":"rileybrown","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/rileybrown.jpg"}},{"id":"b513957e-336d-40c0-a9f6-283bdc80372f","tweet_url":"https://x.com/supabase/status/2096584213512196573","tweet_text":"last person to comment gets free supabase for a year","summary":"last person to comment gets free supabase for a year","author_name":"Supabase","author_handle":"supabase","timestamp":"2026-09-06T13:00:03+00:00","comment_count":993,"like_count":839,"retweet_count":15,"submitted_at":"2026-09-06T13:35:03.15062+00:00","created_at":"2026-09-06T13:35:03.15062+00:00","seo_title":"Supabase Offering Free Annual Plan to Final Comment","bookmark_count":123,"image_url":null,"keywords":["saas","product launch","community engagement","startup growth","developer tools","web development","cloud infrastructure"],"video_url":null,"video_thumbnail_url":null,"image_grid_urls":null,"profile_id":"edc5b08d-6b3b-45a8-96d7-b7759236ddad","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":1,"noindex":false,"profiles":{"id":"edc5b08d-6b3b-45a8-96d7-b7759236ddad","name":"Supabase","handle":"supabase","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/supabase.jpg"}},{"id":"0bc55aca-cf6b-465b-87d1-867a79898eff","tweet_url":"https://x.com/theo/status/2096854674938941448","tweet_text":"If you understand your whole codebase, the codebase isn't that important.\n\nReal software is larger than what your brain can comprehend.","summary":"Understanding architectural principles, design patterns, and system abstractions matters more than memorizing code details because real-world software systems exceed individual cognitive capacity and require mental models that scale beyond line-by-line comprehension. Builders and founders benefit from this mindset by focusing on designing systems with clear boundaries and interfaces, while AI practitioners should optimize for generating code that maintains clarity through structure rather than assuming comprehensive codebase understanding is necessary for effective development.","author_name":"Theo - t3.gg","author_handle":"theo","timestamp":"2026-09-07T06:54:46+00:00","comment_count":24,"like_count":269,"retweet_count":5,"submitted_at":"2026-09-07T07:25:03.250409+00:00","created_at":"2026-09-07T07:25:03.250409+00:00","seo_title":"Understanding Your Codebase Means It's Too Small","bookmark_count":68,"image_url":null,"keywords":["software engineering","programming","system design","code architecture","cognitive load","learning","brain","productivity"],"video_url":null,"video_thumbnail_url":"https://pbs.twimg.com/amplify_video_thumb/2096853248754016256/img/oadTg8Y6i25lhlZc.jpg","image_grid_urls":null,"profile_id":"e0b27d04-f997-4aa1-8b8d-fe85b21c1310","link_card_url":null,"link_card_title":null,"link_card_image_url":null,"link_card_domain":null,"is_pinned":false,"pinned_at":null,"content_type":"tweet","article_title":null,"article_image_url":null,"slug":null,"quality_score":4,"noindex":false,"profiles":{"id":"e0b27d04-f997-4aa1-8b8d-fe85b21c1310","name":"Theo - t3.gg","handle":"theo","avatar_url":"https://dcuhjhnlzlpnshikrfwz.supabase.co/storage/v1/object/public/tweet-media/avatars/theo.jpg"}}],"count":10}