Karan Checker retweeted
No bots. No templates. Just our team crafting your space by hand. HouseKraft is now offering fully bespoke services — designed around you, run personally by real people who care about the details. DM us to start. 🏡
1
2
19
Karan Checker retweeted
Vibe-coded a slick site, then remembered the contact form needs a backend? It doesn't. Point your form at a 000form endpoint: <form action="000form.com/f/…"> Or just ask AI. Submissions land in your inbox. No server, no SDK, no deploy. 000form.com
1
1
8
I’m sure the US govt has an independent AI lab working on its own AI models.
1
57
Karan Checker retweeted
Not only can GPT-6 hear, it can talk! It created a spectrogram in SVG of "Mary Had a Little Lamb"
So Astra is able to identify sounds from mel spectrograms zero-shot. I don't think we've scratched the surface of what this model can do (and this is light reasoning btw)
79
77
1,364
135,789
Yup
Astra can create beautiful three.js UIs entirely in code. You don't need an assets library or MCP to create detailed 3D models. These are all code and are easy to customize for your agent. Give Astra a video, URL, or code reference, then ask it to keep comparing the result against the reference and fixing mistakes. Liquid Glass, shaders, motion design, procedural three.js models, and full environments. All from a few prompts. Some prompts I used: "Recreate this in a single HTML file with three.js. Keep comparing it against the reference and fix any differences." "Add Liquid Glass buttons. On hover, animate the lighting around the borders and make the background gradients move slightly with the pointer. Add subtle particles." "Improve the lighting, edges, and modeling. Give the keys more depth and verify the result against the real MacBook." Too many projects I could open-source. Let me know which one should I release.
1
68
End of captcha?
Astra has successfully beat all 48 levels of the “I’m Not a Robot” game:
2
1
74
Astra can identify sounds from spectrograms
So Astra is able to identify sounds from mel spectrograms zero-shot. I don't think we've scratched the surface of what this model can do (and this is light reasoning btw)
1
1
82
Karan Checker retweeted
Software built for humans is fast becoming infrastructure for agents. The interaction model is shifting from: Human ↔ software to: Human ↔ agent ↔ software Most companies are not ready for this.
41
7
118
11,446
Here's the pattern in AI right now. Every time a stronger model ships, capabilities jump, and people rush to build more ambitious software on top of it. The reports come in fast: this was one-shotted, that tool is obsolete, this whole category is dead. Then, a little while later, the thing they just built gets swallowed too. Some people see the newly possible software and think it's their opening — a chance to ride the wave and stake out the category before anyone else. What they miss is that a capability like this doesn't open a category. It closes one. If the model can build it in a single prompt today, then everyone can build it in a single prompt, and there's nothing left to own. The whole world will work that out soon enough
1
55
Karan Checker retweeted
ORNL researchers developed an AI-guided system that can arrange molecules to build functional materials atom by atom, opening new possibilities for electronics and quantum materials. Read more 👇 bit.ly/4y6QluQ
56
203
1,006
149,673
What’s a good solid audio Sfx ai model?
1
60
Karan Checker retweeted
Introducing Google Pics 🎨 Precise, powerful AI image creation is rolling out now to Google Workspace customers and Google AI Pro & Ultra subscribers. Google Pics lets you edit individual objects, refine or translate text, and collaborate with your team. Try it at pics.new.
286
1,049
10,874
3,051,636
This is amazing.
Fable 5.1 is a beast at many things but one thing in particular I have been having a ton of fun with is generating videos through code. For this one I gave it a picture of a property lot. It designed a house for the lot, rendered it, and produced a cinematic walkthrough.
48
Massive compression of intelligence
We compressed Hy4-preview from 1.5TB to ~200GiB GGUF and it still works well ! Meet MIX-STQ1_0.The trick isn’t just going low, it’s deciding where: calibration data picks each layer’s bit-width, some down to 1.31-bit STQ1_0, some up to 2.06-bit IQ2_XXS. Same budget, lower error. Accuracy barely moves vs BF16 📊 MCP Atlas 83.7→83.2 📊 SWE-Bench multi 82.9→81.3 📊 MRCR 81.3→81.1 📊 IFBench 73.5→72.5 See the details on HF : AngelSlim/Hy4-preview-GGUF Weights & low-bit GGUFs 👇 huggingface.co/AngelSlim/Hy4… #LLM #Quantization #llamacpp #Hy
4
150
One bad experience in a day is enough to derail your brain, suck the positivity out of you, and put you in a bad mood, specially if you’re unwell.
1
1
75
Karan Checker retweeted
Big US labs have absolutely no competitive advantage in flash model range - Luna, $1.2, AA score 52 - Sonnet 5, $10, AA score 55 - Haiku 4.5, $5, AA score 30 - Gemini 3.7 flash, $3.75, AA score 56 Mean while Chinese open models: - GLM 5.3 flash, $0.5, AA score 57 - Qwen 3.8 27B, run locally, AA score 52 - Qwen 3.8 flash next, unknown yet, but will be super cheap and capable - deepseek v4 flash, $0.66, AA score 51 The only way out for OpenAI, Anthropic, Google, and SpaceXAI is keep releasing top models and slim down profit margins with better usage. The competition is keeping them on their toes, and it’s net benefit for users
29
17
472
39,091