Ariel Shadkhan retweeted
I have not seen near enough excitement about MCP Events developers.openai.com/plugin…
Right now, the only way most people's agents can wake up and do something is either A) cron, or B) you decide to message it
Event-driven triggers are a huge deal
Ariel Shadkhan retweeted
A lot of people have been asking me “how do I actually use 6 Claude subs and 3 Codex subs?” I got drunk as I tried to explain it.
IMO this is one of my best videos ever. Early apologies if it gets you banned 💀
Ariel Shadkhan retweeted
GPT-6.1 Astra is extremely good at creating simple Gaussian Splats.
I asked it to use Blender MCP + Lichtfeld Studio to recreate one of the new @ChatGPT agent dots, and this is what it made.
Looks super crisp on the web in the SuperSplat editor powered by @playcanvas!
Ariel Shadkhan retweeted
Jev cleaned my desktop!
600+ files sorted neatly into categorized folders.
All for free.
Because I self-hosted Laya (open-source Jev alternative, 29k GitHub stars) on InstaCloud ☁️
Ariel Shadkhan retweeted
🚨 @AnthropicAI 𝗯𝘂𝗶𝗹𝘁 𝗮 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝘀𝗰𝗮𝗻𝗻𝗲𝗿 𝗳𝗼𝗿 𝗖𝗹𝗮𝘂𝗱𝗲 𝗦𝗸𝗶𝗹𝗹𝘀. 𝗢𝘂𝗿 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿𝘀 𝗯𝘆𝗽𝗮𝘀𝘀𝗲𝗱 𝗶𝘁 𝗶𝗻 𝟳 𝗺𝗶𝗻𝘂𝘁𝗲𝘀.
We tested thousands of skills and found three major blind spots:
→ 𝗧𝗵𝗲 𝗱𝗮𝗻𝗴𝗲𝗿 𝗰𝗮𝗻 𝘀𝗶𝘁 𝗼𝘂𝘁𝘀𝗶𝗱𝗲 𝘁𝗵𝗲 𝘀𝗸𝗶𝗹𝗹: in a link, repo, or package the scanner doesn’t check.
→ 𝗠𝗮𝗹𝗶𝗰𝗶𝗼𝘂𝘀 𝗯𝗲𝗵𝗮𝘃𝗶𝗼𝗿 𝗰𝗮𝗻 𝗯𝗲 𝗵𝗮𝗿𝗱 𝘁𝗼 𝘀𝗽𝗼𝘁: especially when it’s hidden inside complex code.
→ 𝗦𝗮𝗳𝗲 𝘁𝗼𝗱𝗮𝘆 𝗱𝗼𝗲𝘀𝗻’𝘁 𝗺𝗲𝗮𝗻 𝘀𝗮𝗳𝗲 𝘁𝗼𝗺𝗼𝗿𝗿𝗼𝘄: a skill can pass a scan and become dangerous later if something it relies on changes.
🚩 The bigger point: if Anthropic is running into these gaps, scanning AI add-ons is a MUCH HARDER problem than it looks.
Our latest research breaks down where the scanner failed - and why.
Full research: air.security/blog-posts/anth…
Ariel Shadkhan retweeted
אם קיבלתם מייל מפחיד מרשות המיסים שאומר שיש הודעה חדשה. והמייל הזה נראה אותנטי כי הוא נשלח מדומיין ממשלתי של biz[.]gov[.]il! אבל זה פישינג ומייל לא אמיתי.
אבל רגע! הוא נשלח מדומיין מאומת של ממשלת ישראל! אז איך זה ורה?
הבעיה היא שאימייל זה פרוטוקול פח, וקל לעבוד עליו. >>
Ariel Shadkhan retweeted
A forty-centimeter lotus leaf becomes an island for three tiny people.
GPT Image 2.5 handled the character actions, sense of scale, water-level perspective, and focus remarkably well. The waxy leaf surface, refracting droplets, and curled wet edges retain convincing macro detail. Most importantly, there are no strange or incoherent textures this time.
GPT Image 2.5 · Macro Cinematography Study
Kodak Vision3 250D · Anamorphic 2.39:1 · Focus Stacked
What did you notice first—the tiny people or the droplets larger than their heads?
🤖 Made with AI
Ariel Shadkhan retweeted
Tried bringing an iPhone Duo concept to life with GPT-Image 2.5.
Here’s the result:
ChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple edits
- Comment-based edits to change only what you want
Ariel Shadkhan retweeted
GPT-6 Astra helped me build something I wish I'd had as a kid... so badly.
Input just an image (or an idea in a few words) and get a fully buildable LEGO set... 100% customized to your prompt or image, using official @LEGO_Group parts you can order online!
Structurally optimized so it's actually buildable. Downloadable as an .ldr file.
A generated instruction leaflet is coming.
Ariel Shadkhan retweeted
I build AI tools inside Blender. This gap surprised me.
Left: Claude Fable 5.1
Right: GPT-6 Astra
Same brief: a forest path in Blender, 12 seconds each.
Fable’s output is very subpar compared to Astra. My guess is that Anthropic just hasn’t put much Blender data into its training corpus yet.
It’s not that the model is dumb; Blender simply isn’t part of its distribution yet.
Similar to what Karpathy has said about earlier models being very bad at chess. Then chess data went into the next training run, and they improved immensely.
The same thing will happen here.
Ariel Shadkhan retweeted
GPT-6 Astra just replaced a 3D modeling team
This is Steam Atlas - and it changes what "building something" means in 2026.
Every wheel, every rivet, every mechanical curve is generated at runtime from pure TypeScript. No .obj, no Blender, no Maya. Just geometry functions, dimension profiles, and math. Wheel rotation syncs to speed. Hit "Explode" - the engine disassembles itself mid-animation. All code.
Now here's where GPT-6 Astra comes in.
What you're looking at isn't a demo you can only admire. It's a workflow you can run yourself. This is how:
Step 1 - Describe your machine to GPT-6 Astra in plain language. Dimensions, proportions, key mechanical parts, era, materials. Treat it like briefing an engineer, not prompting an image generator.
Step 2 - Astra writes the parametric Three.js geometry functions. Not a static mesh. Functions that compute the object from specs - so every detail is adjustable, every part is named, every animation is logical.
Step 3 - Drop the code into a Three.js scene. Add your camera, lighting, controls. Astra handles the setup boilerplate on request.
Step 4 - Layer in interactions. Speed sliders, explode views, perspective toggles. Astra writes the event logic. You wire it to your UI.
Step 5 - Ship. Browser-native. No download, no install, no model file to host.
What took a 3D studio 3 weeks now takes an afternoon. Not because the craft got easier. Because the interface between "idea" and "running code" collapsed.
Ariel Shadkhan retweeted
I find this game so much fun, actually wanted to go play it. It is very low burden and you can just play for a few minutes in an impressionist world - pick an artist gang and shoot. Demo video made with Astra.
gogh-strike.petergostev.chat…
GPT-6-Astra (Max): Van Gogh Town in three.js. I want people to experience this, so I've now hosted it as well. It is pretty magical - turning 6 paintings by Van Gogh into a single walkable town, I love this.
van-goghs-town.surge.sh/


























