Introducing Mixedbread Wholembed v3, our new SOTA retrieval model across all modalities and 100+ languages.
Wholembed v3 brings best-in-class search to text, audio, images, PDFs, videos...
You can now get the best retrieval performance on your data, no matter its format.
new Field Theory 0.1.92
> hot mic is now extremely accurate
> new dynamic island
> new parakeet transcription engine (use it!)
> shift+cmd+k adds window / comp control
> lots of QOL
all functions are local / offline. all accounts are free
have fun being iron man
https://t.co/GeD0UJOQQO
@Dimillian It either goes full Neo and tries to rewrite the game state or it is very slow because the agentic loop + path finding + its map visibility are poor.
So mostly I am driving and it just monitors the game state changes and provides commentary and tips.
I've had this emulator copilot running for a couple weeks - was inspired today to add the chat codes and let the copilot edit the game states actually.
Now it can help you catch all the starters. I (claude) could not get teleporting to work but that's what Fly is for ๐ฆ
GPT-5.4 built this for me in 3 prompts.
It hacked the NES Mario ROM to expose RAM events, then created a JS emulator that could send browser requests so every character in the game is controlled by an AI lmaooo
If you are playing with ROMS this weekend, then this is a fun project:
FRLG + EmulatorJS + Poll game state for achievements, milestones, basically everything + Bulbapedia + Claude Code
AI friend for 1P game... can congratulate you on your wins, give advice, help you catch the 2 starters you didn't choose, and gift you infinite nuggets in your PC ๐
Retro achievements is a really cool project. It runs in emulators and polls for game state โณ to add achievements to any game.
Human curated hex codes. Also a great starting point for an agent to extract MORE events so it can track any progress in any game - not just milestones
https://t.co/VRlf4NRd7I
@andyhattemer Not to mention how bad this makes the poster look - both products are good and have their own selling points. Anyone that canโt get it to accept writes should not / would not advertise their own stupidity in this way. Might as well say โpay you $500 to tweet how dumb you areโ
Seedance 2.0 can generate beat-synced videos from audio input.
I uploaded one MP3. This is what came out ๐
The model reads the rhythm + tempo of your track, then auto-syncs:
- outfit transitions on every beat drop
- camera cuts aligned to the music
- motion + pose changes that hit the downbeat
- No manual keyframing. No editing. One prompt + one audio file.
How I did it:
1. Uploaded reference images (character + base outfit)
2. Uploaded an MP3 track as @ audio1
3. Prompted: "Fashion transformation video. The model transitions through 5 distinct outfits, each change synced precisely to the beat drop. Confident runway walk, dynamic camera angles shift with the rhythm. Studio lighting, editorial fashion photography style."
This is the feature that separates Seedance 2.0 from everything else โ audio isn't just background, it drives the entire generation.
@TessaDoesThings The point of the class is to learn the skills in the course description ๐คฆ to go on and take the next class. these replies โkids find it more engagingโ like why not just throw on a good movie then??
Vibe coding deserves its own class ๐
Each of these are pretty accessible through discrete APIs.
Lots of models to load and unload - would slow down throughput and be more expensive to use the unified workflow. Better to batch so each step runs on warmed gpus. Unified workflow is great for local though if youโve got the setup for it ๐
@lucatac0 Vercel supports this too - can serve some paths on framer or wordpress or different vercel deployments - wherever.
Does cloudflare require the domain to be managed in cloudflare still? Only reason we donโt move more things to them are some odd restrictions around that.
@iruletheworldmo Would love to see it showing a few words at a time. If it could do that intelligently even better. Fast readers donโt read word by word.