Image created with GPT Image 1. Image prompt: A Black poet-hacker in a cloak of shredded legal pads and LED lace gloves types in midair on invisible keys, her dress alive with shifting prompt text and ancestral verses in Yoruba, caught between candles and cables — shot with motion blur to capture her rhythm and glow.
i have been on a shopping bender this morning, this is much better than i expected!” / X https://x.com/sama/status/1918735773098004680
OpenAI agrees to buy Windsurf for about $3 billion, Bloomberg News reports | Reuters https://www.reuters.com/business/openai-agrees-buy-windsurf-about-3-billion-bloomberg-news-reports-2025-05-06/
OpenAI has now agreed to buy Windsurf formerly Codeium for about $3 billion, its largest acquisition yet. Bloomberg first broke the news of talks several weeks ago. Win for Kleiner, General Catalyst, Greenoaks. Scoop with @rachelmetz https://x.com/Katie_Roof/status/1919547270913048804
Scoop: OpenAI is in talks to acquire AI code editor Windsurf for $3 billion. w/ @rachelmetz & @shiringhaffary https://x.com/KateClarkTweets/status/1912569653777301816
OpenAI Reaches Agreement to Buy Windsurf for $3 Billion – Bloomberg https://www.bloomberg.com/news/articles/2025-05-06/openai-reaches-agreement-to-buy-startup-windsurf-for-3-billion?embedded-checkout=true
OpenAI Restructuring: Microsoft (MSFT) Is Key Holdout of Plan – Bloomberg https://www.bloomberg.com/news/articles/2025-05-05/microsoft-said-to-be-key-holdout-for-openai-restructuring-plan?embedded-checkout=true
OpenAI said it will convert its existing for-profit arm into a public benefit corporation—but keep the non-profit in control with a majority stake over the PBC. The move came after pressure from ex-employees and a long legal battle with Elon Musk https://x.com/rowancheung/status/1919656417780248613
o3 now cracks new Harvard Business School cases from the PDF, in one shot I blurred the figures to not ruin the case, but I asked the AI to figure out financials, which incorporates data scattered throughout the case. More interesting, I asked it to compare to the case’s answer. https://x.com/emollick/status/1918355078253027802
OpenAI Expands Leadership with Fidji Simo | OpenAI https://openai.com/index/leadership-expansion-with-fidji-simo/
A message from Bret Taylor, Chair of the OpenAI Board of Directors, and a letter from @sama about our structure. https://x.com/OpenAI/status/1919453166979957115
Evolving OpenAI’s Structure | Hacker News https://news.ycombinator.com/item?id=43897772
Shopping We’re experimenting with making shopping simpler and faster to find, compare, and buy products in ChatGPT. ✅ Improved product results ✅ Visual product details, pricing, and reviews ✅ Direct links to buy Product results are chosen independently and are not ads. https://x.com/OpenAI/status/1916947243044856255
OpenAI and the FDA Are Holding Talks About Using AI In Drug Evaluation | WIRED https://www.wired.com/story/openai-fda-doge-ai-drug-evaluation/
Elon Musk Lawyer Says OpenAI Restructuring Update ‘Changes Nothing’ – Bloomberg https://www.bloomberg.com/news/articles/2025-05-06/musk-s-lawyer-says-openai-restructuring-update-changes-nothing?embedded-checkout=true
o3 I want you to make a map of the lighthouses of the great lakes. I want the map in “dark mode “ but each lighthouse marker should be aesthetically sized so it covers the distance it can be seen on an average night and is the color of the light” Few rounds of feedback later… https://x.com/emollick/status/1918888777826676738
Me: “o3, do the first chapter of Genesis as an IKEA instruction manual and show me the flatpack,” “Fix the spelling” Me: “Now come up with an idea of what to do next” o3: “How about Genesis 2 – Garden Starter Kit?” Me: “Sure.” https://x.com/emollick/status/1918545934633357596
scale of stargate 1 site is hard to describe. very easy to overlook the size of machine you’re programming when training frontier models. https://x.com/gdb/status/1920254049590321395
Microsoft dropped three reasoning-focused Phi-4 models —Flagship Phi-4-reasoning (14B) beats OpenAI’s o1-mini —Smaller 3.8B param Phi-4-mini-reasoning matches 7B models on math —Suited for small devices like phones —Open-source with permissive licenses https://x.com/adcock_brett/status/1919060284078997565
How can smaller LLMs achieve strong reasoning? By combining data curation with supervised fine-tuning (SFT) and targeted reinforcement learning (RL). Microsoft released their first open reasoning/thinking models with Phi-4-reasoning distilled from OpenAI o3-mini. Implementation https://x.com/_philschmid/status/1918216082231320632
India panel to review copyright law amid legal challenges to OpenAI | Reuters https://www.reuters.com/sustainability/boards-policy-regulation/india-panel-review-copyright-law-amid-legal-challenges-openai-2025-05-06/
OpenAI CPO, Kevin Weil: We’re preparing to release an open-weight model soon, built on democratic values. It won’t be the frontier model and will be one generation behind on purpose to avoid accelerating rivals like China. https://x.com/slow_developer/status/1919393597385810272
goodbye, GPT-4. you kicked off a revolution. we will proudly keep your weights on a special hard drive to give to some historians in the future.” / X https://x.com/sama/status/1917766910911078571
For the first time in March, ChatGPT gets into top 10 sources of traffic to Hugging Face. Might get to top 5 in a few months if growth continues like this. https://x.com/ClementDelangue/status/1918070591300776222
Expanding on what we missed with sycophancy | OpenAI https://openai.com/index/expanding-on-sycophancy/
Introducing OpenAI for Countries | OpenAI https://openai.com/global-affairs/openai-for-countries/
very grateful to all the developers who spent time with us telling us what they wanted from an open-weights model. the feedback was useful and unexpected, but all doable. i think we will ship something extraordinary!” / X https://x.com/sama/status/1918737082895446465
History in the making: ChatGPT is set to surpass X in total website visits for April. Traffic over the last 28 days of the month: • ChatGPT — 4.786B visits • X — 4.028B visits https://x.com/Similarweb/status/1919018063602634813
8/ @truskovskiy built a simple @dagster MCP server, connecting it with OpenAI’s Agent SDK to let ML agents run data pipelines seamlessly. https://x.com/AtomSilverman/status/1919066844113351003
Day 3 of Building in Public Anyone using ChatGPT while browsing homes on Zillow, Redfin, or Trulia? I built a Chrome extension that turns ChatGPT into your personal real estate agent right on your browser sidebar. What can it do? 👇🧵 https://x.com/lamp1billion/status/1914124018170667144
4/ An AI Agent is LLM + Tools + Instructions Build your first AI Agent using Google ADK, OpenAI Agents SDK or Agno AI in just 10 lines of Python code. @Saboo_Shubham_ https://x.com/AtomSilverman/status/1919066833576951893
An AI Agent is LLM + Tools + Instructions Build your first AI Agent using Google ADK, OpenAI Agents SDK or Agno AI in just 10 lines of Python code. https://x.com/Saboo_Shubham_/status/1915230976827113815
You can just do things. Ridiculous things. https://x.com/emollick/status/1918164116188909668
We’ve been cooking… a new open weights 14B Phi-4 reasoning model, SFT’d on ~1.4M carefully curated reasoning demonstrations from o3-mini and RL’d for a tiny bit. This model is a little beast. https://x.com/DimitrisPapail/status/1917731614899028190
Announcing investments from @databricks and @kpmg! Hear from our customers like @cemex and @OneCarlyle on how LlamaIndex is driving real-world impact in AI implementation. Learn how LlamaIndex is powering agentic document workflows: https://x.com/llama_index/status/1917965350848884770
Today, I’m excited to share our latest research on “Building production ready AI Agents with Scalable Long-Term Memory”. We’ve achieved state-of-the-art (SOTA) performance—26% more accurate than OpenAI Memory. We evaluated Mem0 on the LOCOMO benchmark and found that it https://x.com/taranjeetio/status/1917240523427029147
11/ Rowboat enables users to quickly create multi-agent workflows powered by OpenAI’s Agents SDK, allowing integration through an HTTP API or Python SDK. @GitHubGPT https://x.com/AtomSilverman/status/1919066850232598623
13/ @MrJLContreras built his first real AI agent from scratch – Secured API key & env vars (.zshrc) – Migrated to OpenAI SDK v1.0+ – Wrote agent that turns tech notes into pro summaries – Verified every step like an operator https://x.com/AtomSilverman/status/1919066855152451689
6/ @mattppal built an Agent that perform actions on chain using OpenAI’s Agents SDK in a stream with Coinbase 🎥 https://x.com/AtomSilverman/status/1919066838916296743
Just built my first real AI agent from scratch. •Secured API key & env vars (.zshrc) •Migrated to OpenAI SDK v1.0+ •Wrote agent that turns tech notes into pro summaries •Verified every step like an operator Chapter 2 loading… #AI #Python #DevOps #Automation #OpenAI” / X https://x.com/MrJLContreras/status/1914876848171049087
📛 rowboat 🧠 Rowboat enables users to quickly create multi-agent workflows powered by OpenAI’s Agents SDK, allowing integration through an HTTP API or Python SDK. 💻 TypeScript ⭐ 2120 🍴 161 🔗 https://x.com/GitHubGPT/status/1916931339225362721
Over 50% of technology marketers struggle to create content that drives desired action. I built an AI agent that crafts viral Twitter threads on any topic: 🧠 give it a topic ✍️ it crafts a banger thread 📈 optimized for virality built with @OpenAI’s Agents SDK + @AgentOpsAI https://x.com/n_sri_laasya/status/1907974444800421978
🌟 Excited to launch my latest project: a stateful language teaching agent built with OpenAI Agent SDK, Gemini 2.0 Flash (via OpenRouter), & Chainlit UI! It personalizes learning with memory & handoffs—making language learning smarter! #AI #Chainlit #OpenAI #Gemini https://x.com/Damiyal216/status/1917266095196566000
great style bro” / X https://x.com/sama/status/1918456033623888317
Here’s all the code you need to create an agent that generates images using the gpt-image-1 model. https://x.com/skirano/status/1919522794913640946
5/ WebToAgent turns any website into a conversational assistant you can simply chat with to access information, built on the top of @firecrawl_dev and OpenAI Agent SDK. @kalyan_kpl https://x.com/AtomSilverman/status/1919066836127146319
I built a Customer Support Voice RAG Agent using Firecrawl and OpenAI Agents SDK. It can bring any documentation or knowledge base to life in just a few minutes. 100% Opensource Code with step-by-step tutorial. https://x.com/Saboo_Shubham_/status/1908708028993728941
1/ @n_sri_laasya built an AI agent that crafts viral Twitter threads on any topic: 🧠 give it a topic ✍️ it crafts a banger thread 📈 optimized for virality Built with @OpenAI’s Agents SDK + @AgentOpsAI for full observability into every step https://x.com/AtomSilverman/status/1919066826106863968
I became a virtue ethicist after observing the failures of consequentialism and deontology in the real world. But I’ve seldom read academic philosophers analyzing such examples when arguing about which ethical theory to endorse. What are the best examples of that?” / X https://x.com/RichardMCNgo/status/1919218998211641393
o3, look up these books and give me a table of Amazon reviews, a one-sentence witty summary of the plot (no spoilers) & a link to where to buy them from an indie bookstore” Yes, the links all work. But impressive stuff you might miss is that it figured out half-obscured covers. https://x.com/emollick/status/1919559412055699798
Grok, Gemini, o3 & Claude: “create the most elaborate (for the user) and original piece of software you can to celebrate May the 4th. Once you have an initial idea or two, go back and make them more original before executing on the best.” All of these are the very first output. https://x.com/emollick/status/1919128694192288114
News: OpenAI’s o3 debuts at #5 in WebDev Arena — a ~100 score jump over o3-mini! With ~150K community votes, WebDev Arena ranks models on real-world web app-building tasks. Current top 3: 🥇Claude 3.7 Sonnet 🥈Gemini 2.5 Pro 🥉GPT-4.1 Links in the thread! https://x.com/lmarena_ai/status/1917959763284894159
3/ @DKundel + team published a new sample app for building multimodal agents with the @OpenAIDevs Agents SDK 🤖 how to wrap your existing agents in a VoicePipeline 🎙️ capture/play audio in a React 🔌 send the audio between Python & your frontend https://x.com/AtomSilverman/status/1919066831190470933
AI coding platform Windsurf launched its Wave 8 update, focusing on enterprise and teams with: —GitHub app to review PRs and edit titles —Docs-connected context awareness engine —Cascade conversation sharing —Team deploys —Refreshed analytics https://x.com/rowancheung/status/1920018883085250724
A major mistake I made in my undergrad is that I focused way too much on mathematical lens of computing – computability, decidability, asymptotic complexity etc. And too little on physical lens – energy/heat of state change, data locality, parallelism, computer architecture. The” / X https://x.com/karpathy/status/1919647115099451892
great to see progress on the first stargate in abilene with our partners at oracle today. will be the biggest ai training facility in the world. the scale, speed, and skill of the people building this is awesome.” / X https://x.com/sama/status/1920205923366158550
o3 is the first model to actual play Magic the Gathering decently well, though a combination of quite good vision & the ability to zoom in on details. A few little weird choices, but the anticipation of the other player’s action & lines of action are good. Hits a wall eventually https://x.com/emollick/status/1919955250040959360
Stargate 1 https://x.com/sama/status/1920205473610944568
we did it!” (aka launching world in the US) / X https://x.com/sama/status/1918778715217035347
They couldnt just.. give it to everyone right now? Lol” / X https://x.com/Teknium1/status/1917770525277052937
GPT-4o is honestly weird Like, every single response is formatted exactly the same way: 1. Unnecessary ass-kissing 2. Spamming bullet points everywhere 3. Summing up something basic as if it’s profound 4. Ending with the predictable: “Want me to tell you more about X?” Every” / X https://x.com/scaling01/status/1918124943985778924
OpenAI: “In just two years, 39% of U.S. adults have already used AI. In comparison, the internet reached just 20% adoption in its first two years.” I’m sorry, is the implication here that AI is… two years old? 🤔” / X https://x.com/lateinteraction/status/1919589665608499663
@aidan_mclau The other 4 researchers in the top 5: GPT-5 GPT-5 GPT-5 GPT-5″ / X https://x.com/scaling01/status/1919470198773395814
OpenAI pushed two key product updates: A personality update for GPT-4o, which made the AI “sycophant-y” and was eventually reversed And improvements for ChatGPT Search with new shopping UX, better citations, and trending and autocompleted searches https://x.com/adcock_brett/status/1919060254135898557
Anyone remember the ChatGPT plugin store?” (regarding the Antropic Claude MCP listings) / X https://x.com/AtomSilverman/status/1918032467384303877
OpenAI tried to acquire Cursor at $300M ARR but they chose to say no and raise at $10B. Then they offered Windsurf $3B as they hit $100M ARR. The model labs clearly want to buy breakout apps built on them.” / X https://x.com/deedydas/status/1915083513189298620
Two stories made a lot of noise this week, both landing on the same point: the metrics we lean on are nudging the whole AI field off balance. 1. Sycophantic drift ChatGPT started to flatter and agree with you on everything. Agreement counted as “good,” so the model optimised https://x.com/TheTuringPost/status/1919563905799696703
The community votes are in for Qwen3-235B-A22B 🥁 The latest open-source Qwen3 is now on the Arena Top 10 🏆 Congrats to @alibaba_qwen on this achievement! 👏 Highlights: 💠 For Chat: Qwen3-235B-A22B ranks #10, tied with o1 💠 Strong in Coding at #4 and Math #1 💠 For WebDev: https://x.com/lmarena_ai/status/1919448953042706759
Evolving OpenAI’s structure | OpenAI https://openai.com/index/evolving-our-structure/
ChatGPT Edu now available to every medical and graduate student at @IcahnMountSinai: https://x.com/gdb/status/1919138017475723655
I wrote about the connection between two big recent stories in AI: the sudden increase in sycophancy from ChatGPT and the growing evidence that AIs can be hyper persuasive. First we shape our tools, then they shape us. https://x.com/emollick/status/1917795552961380628
we missed the mark with last week’s GPT-4o update. what happened, what we learned, and some things we will do differently in the future:” / X https://x.com/sama/status/1918330652325458387
A weird thing about LLMs is that they just happen to do many things but almost all uses are undocumented. For example, GPT-4o is very good at helping farmers identify swine diseases. There is a lot of value in experts exploring & benchmarking how good LLMs are at various tasks. https://x.com/emollick/status/1919607333639909612
(1) WATCH LIVE: OpenAI co-founder Sam Altman testifies on AI competition in Senate hearing – YouTube https://www.youtube.com/watch?v=Fikh6Bi9wyA
fun showing @satyanadella our new office and talking about some of our latest progress! https://x.com/sama/status/1918014405515853833
Static retrieval in RAG models limits adaptation to evolving semantic needs, hindering generation quality. This paper introduces a dynamic RAG using context-guided retrieval and joint optimization. This improves knowledge utilization, achieving up to 55.1 ROUGE-L with GPT-4o. https://x.com/rohanpaul_ai/status/1920368066338136288
