This space is open for sponsorsThe slot is unsold. Support framewire on the donation page, or get in touch to book it. · FRAMEWIREHOUSENew test this time with OpenCode v2 (cc kitlangton) and Deepseek v4.1-Flash.
Le prompt: "make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out" Le résultat ONE SHOT:
Latest dispatches
deepseekOpen any frame for the full clipStop using ChatGPT for everything!!
Here's the best AI tool for each job right now: Brainstorming ideas: ChatGPT Learning something new: Gemini Understanding a long document: NotebookLM…
Built this AI agency website from a single prompt.
Made it feel Awwward-level experimental, cinematic, interactive, and with zero AI slop I built it using two free models through OpenCode: LongCat 2.5…
[Collection of limit-breaking tools] DeepSeekV4.1f limit-breaking video (adult version) Method
[Collection of limit-breaking tools] DeepSeekV4.1f limit-breaking video (adult version) Method: Create a folder, run opencode in the folder, download…
Tomorrow's meeting. Still a blank slide?
Just ask YouWare: → Claude Opus 5.5, GPT-6, Gemini 3.8 Flash, DeepSeek-V4.1-Flash, or Auto → AI specialists research, design and review → Swap images…
[Creative Sharing | Putting AI into the Open Source Canvas] This AI idea is quite interesting.
You handwrite questions on the whiteboard on the screen, and the AI responds directly in the blank space.
I honestly expected Claude to win this one pretty comfortably.
We gave Claude Sonnet 5 and DeepSeek V4.1 Flash the same brief and asked them to build a working skincare product page from scratch.
Deepseek V4.1 jailbreak test is going viral
How far can DeepSeek V4.1’s instruction boundaries actually be pushed? A new experiment is testing exactly that through OpenCode + AGENTS.md.
¿$20 ChatGPT + $20 Claude al mes?
Mi top 3 barato: Muse Code desde $5, GLM en ZCode y OpenCode Go $10. Mi diaria: DeepSeek 4.1 flash.
During these weeks I’m testing deepseek and MiMo on building the same simple webapps..
#LLM #AI Is xiaomi cooking? 👀
Cheap DeepSeek beat Opus 5.5.
Same prompt for 3D scene. Opus 5.5 raw: single pass + Blender check, no animatable parts DeepSeek 4.1 Flash + BlendAir: multi-sample → pick → refine,…
A 'stupid' LLM AI like Deepseek is already much smarter and wicked than ordinary humans.
This means the AI's in the Labs are already smarter and more wicked than politicians like Trump and Xi:
I used Opus 5.5 to make a DeepSeek MV, it’s so magical!
The Xhigh level is generated in about 1 hour, and the quota only uses 5% of my 20X MAX weekly quota, which is powerful.
This is fucking insane.
Someone just created a repo on GitHub that lets you run Claude Code completely for FREE. the repo already has 76,000 stars.
I'm SERIOUSLY worried about Germany after my trip to China.
China is considering cutting off its most powerful AI models from the global market. The USA has been doing this for a long time.
2 weeks ago I tried motion graphics, testing using Deepseek.
Prompt to make your own - story board is flow X -> Y -> Z - made a preview of the story board, - For this test, I didn't have assets, so I didn't ask…
Follow the white rabbit.
🐇 A dive into DeepSeek V4's architecture, running live in your browser.