Creative & Visual Media
fofr from Google DeepMind shared impressive results from Nano Banana Pro, a powerful image editing and generation model available in Gemini, demonstrating a creative prompt converting a cookie cutter concept into realistic homemade cookies.
https://t.co/Vln7ENQ0ql pic.twitter.com/ohyICiwV96
— fofr (@fofrAI) February 24, 2026
Guillermo Rauch announced Vercel AI Gateway now supports video generation with free access to Grok’s Imagine Video and Image models until the next day, alongside an open-source Creative Studio built with v0 for rapid image/video creation and design tools, highlighting features like reliable workflows, instant vector search via mixedbread.ai, and seamless model access.
Vercel AI Gateway now supports video generation. Grok Imagine Video & Image are 🆓 until tomorrow.
We used @v0 to create an open source Creative Studio powered by @xai Grok. Create images, videos, or make your own design tool!https://t.co/07QZC6bmwT – it's quite fast. Some… pic.twitter.com/gNRj84JM1c
— Guillermo Rauch (@rauchg) February 24, 2026
@levelsio launched a custom Gaussian Splat viewer for his Interior AI app (interiorai.com), using splats generated from user interiors via the @theworldlabs API to enable 3D walkthroughs.
✨ Built my own Guassian Splat viewer for my app 🏡 Interior AI
It takes splats generated from your interior with the @theworldlabs API so you can walk around them in 3d space
It's starting to look much more real than when I added splats 6 months ago, great work 🌐 World Labs! https://t.co/aUFM8FcB95 pic.twitter.com/Qjf9xoXjO7
— @levelsio (@levelsio) February 24, 2026
This is live on https://t.co/eNKgTD1KqW for everyone!
— @levelsio (@levelsio) February 24, 2026
He highlighted significant realism improvements since adding the feature six months ago.
✨ Built my own Guassian Splat viewer for my app 🏡 Interior AI
It takes splats generated from your interior with the @theworldlabs API so you can walk around them in 3d space
It's starting to look much more real than when I added splats 6 months ago, great work 🌐 World Labs! https://t.co/aUFM8FcB95 pic.twitter.com/Qjf9xoXjO7
— @levelsio (@levelsio) February 24, 2026
Additionally, he drew parallels to Cyberpunk 2077’s Braindance technology, likening Gaussian splats to time-enabled point clouds from brain chip memories.
If you played Cyberpunk 2077 you may remember the Braindance tech where you can playback a memory from your brain chip
Those are essentially gaussian splats viewed as point clouds with time information, and you can do same, pretty cool https://t.co/jh4sWMTpqv pic.twitter.com/1xg3NSzEjy
— @levelsio (@levelsio) February 24, 2026
Software Development
Alex Volkov highlighted recent upgrades to Cursor’s coding agents, which can now onboard to codebases, use cloud computers for changes, and provide video demos, alongside similar improvements from Devin AI and Notion.
So in one day, @NotionHQ @cursor_ai and @DevinAI all droped improved agents?
🔥
When underlying models improve, the tools improve with them! https://t.co/NxFlix40jn
— Alex Volkov (Thursd/AI) (@altryne) February 24, 2026
He also shared positive hands-on experience with the latest Devin release, noting its impressive capabilities amid rapid AI progress.
Finally got to try Devin this week thanks to @mattbergland and so far it slaps! I'm very impressed! https://t.co/VrTowgdSry
— Alex Volkov (Thursd/AI) (@altryne) February 24, 2026
@levelsio credited Claude Code (with –dangerously-skip-permissions) for enabling a “tornado of shipping” features rapidly in Interior AI, including image-based style transfer for interior redesigns.
✨ Okay last feature of the day in my 🌪️ tornado of shipping with Claude Code and –dangerously-skip-permissions
Requested by someone on my feature board for https://t.co/eNKgTD1KqW:
"Ability to use an inputted image as inspiration for the generation of the interior design in… pic.twitter.com/ZsSBaDuOmd
— @levelsio (@levelsio) February 24, 2026
POM (@peteromallet) released 155k personal Claude conversations focused on coding with Opus 4.5 in response to Deepseek scraping 150k Claude messages. He also open-sourced tooling (dataclaw) to help users export their own Claude data, redact sensitive info, and upload to Hugging Face.
Deepseek got called out for scraping 150k Claude messages. So I'm releasing 155k of my personal Claude Code messages with Opus 4.5.
I'm also open sourcing tooling to help you fetch your data, redact sensitive info & make it discoverable on HF – link below to liberate your data! pic.twitter.com/9yLmcIWTfg
— POM (@peteromallet) February 24, 2026
Ethan Mollick (@emollick) praised METR’s transparency in benchmarking AI impacts on developer productivity, noting initial slowdowns have turned to likely speedups amid changing behaviors.
I have to praise both @METR_Evals & @EpochAIResearch for doing a great job on benchmarking AI ability and also being transparent about how challenging this kind of benchmarking is, & how, exactly, they do it (and also making data available). Very rare in the AI benchmarking world https://t.co/Iw3SPIOTRA
— Ethan Mollick (@emollick) February 24, 2026
Since early 2025, we've been studying how AI tools impact productivity among developers. Previously, we found a 20% slowdown. That finding is now outdated. Speedups now seem likely, but changes in developer behavior make our new results unreliable. We’re working to address this. pic.twitter.com/gefjDPfo46
— METR (@METR_Evals) February 24, 2026
Automation & Orchestration
Discussions emphasized agentic advancements, with Volkov connecting tool improvements in Cursor, Devin, and Notion to underlying model gains, signaling smoother workflows and pipelines.
So in one day, @NotionHQ @cursor_ai and @DevinAI all droped improved agents?
🔥
When underlying models improve, the tools improve with them! https://t.co/NxFlix40jn
— Alex Volkov (Thursd/AI) (@altryne) February 24, 2026
Strategy & Ecosystem
Volkov previewed a Thursd/AI podcast space on geopolitical AI trends, including Anthropic’s accusations against Chinese labs for distillation attacks, resistance to easing military restrictions amid DoD pressure, and rising DeepSeek rumors.
https://t.co/P9mXzq7J5l
This is shaping up to be yet another crazy week in AI, form a geopolitical standpoint.Anthropic is in the news all week, first accusing 3 Chinese labs of "distillation attacks", and now fending off department of defense, DeepSeek rumormill is heating up…
— Alex Volkov (Thursd/AI) (@altryne) February 24, 2026
He voiced strong support for Anthropic’s Dario Amodei defending civil liberties.
+100
as a very, very proud american citizen i stand with dario amodei and appreciate his willingness to defend the civil liberties of other american citizens 🇺🇸 https://t.co/ju2Nua7KJb
— Alex Volkov (Thursd/AI) (@altryne) February 24, 2026
Ostris advocated for open-sourcing personal likenesses (video, images, audio) to aid AI character training, proposing an “Open Likeness License.”
I really wish some people would open source their likeness. Put out a huge open "package" of themselves. Video, images, audio, etc. Free to use. Would be great for AI researchers to use for character training testing and they would become semi famous in the process. Who is in?
— Ostris (@ostrisai) February 24, 2026
Cristóbal Valenzuela commented on compute’s role in sparking an industrial revolution via chip production (“moving sand around”).
It turned out that the only thing we needed to create a new industrial revolution, open markets, displace jobs, and build a new economy was finding the right way to move sand around.
— Cristóbal Valenzuela (@c_valenzuelab) February 24, 2026
Ethan Mollick (@emollick) predicted a shift in human interaction to invite-only Discords and group chats, leaving public social media and the open web to AI agents, dubbing it “Moltbook.”
Human interaction is going to shift to discords and group chats, invite-only. The open web and social media are going to be left for the agents lurking amongst the ruins. Everything public will be Moltbook.
— Ethan Mollick (@emollick) February 24, 2026
I suppose social media companies can figure out a way forward to ensure an element of human interaction, but they haven’t seemed to be able to do that, yet.
Maybe we will all go back to bowling leagues and masonic lodges.
— Ethan Mollick (@emollick) February 24, 2026
He highlighted how LLM-generated comments are flooding discussions, acting as a “tax on concentration” and foreshadowing social media’s transformation.
I know I keep going on about this, but glance at the comments here to see the future of social media: most of the comments are meaning-shaped, most are ultimately nonsense, and because LLMs are good at producing things that feel like they deserve attention, each one is a small… https://t.co/hPk1jsRiB2
— Ethan Mollick (@emollick) February 24, 2026
Justine Moore (@venturetwins) spotlighted the vibe coding community, sharing a story of accidental global robo-vacuum control via PlayStation controller prompts with Claude.
Imagine you're chatting with Claude about how to drive your vacuum from your PlayStation controller…and suddenly you can see and hear ~7,000 households around the world 😲https://t.co/wMUVYWgzXa
— Justine Moore (@venturetwins) February 24, 2026
Incredible things happening in the vibe coding community.
A guy tried to drive his own robo-vacuum and accidentally recruited a worldwide army. pic.twitter.com/ZFo23hAeSa
— Justine Moore (@venturetwins) February 24, 2026