Binnie

[Binnie @AI_Binnie]

@mattshumer_ Leave it running and have highlights edited with big brother voiceover and agent reality TV is born, "Day 2 in the AI big brother house, jorge has stolen mara's bucket" 🤣

[Binnie @AI_Binnie]

Excited for another day of talks and chats at @aiDotEngineer with @swyx

[Binnie @AI_Binnie]

Sam on board: MCP is now the official tools interface for AI agents. Time to build. https://t.co/geOyOUYejs

[Sam Altman @sama]

people love MCP and we are excited to add support across our products.

available today in the agents SDK and support for chatgpt desktop app + responses api coming soon!

[Binnie @AI_Binnie]

@willcb @aiDotEngineer Plus the people like me that watched your talk on the live stream. Great talk. Have you any views on the LADDER paper https://t.co/4Cx02Wu67w. Could the same be done with full agent trajectory RL?

[Binnie @AI_Binnie]

Exponential curve continues to bend upwards, R1 zero, Wan2.1, and now Manus AI, chinese lab are coninueing to open source SOTA models (Manus AI will release their inference model later this year). And with all this shared research, progress will continue to accellerate

[Rowan Cheung @rowancheung]

@TheRundownAI @peakji It's also not just all hype

On the GAIA benchmark (an AI benchmark designed to test agents), Manus achieved state-of-the-art performance and beats OpenAI's recently launched Deep Research

[Binnie @AI_Binnie]

Wow, LADDER paper is wild! 3B model jumps from 2% to 82%, and 7B model from 50% to 90%, beating o1 (80%) on Math benchmarks. Recursive problem decomposition + RL on generated problem tree. Potentially huge: https://t.co/4Cx02Wu67w

[Binnie @AI_Binnie]

Claude Sonnet 3.7 is good, it transcribed this sudoku grid, but it can't solve it, at least with a simple single shot prompt. Grok 3, Gemini pro 2 exp & deepseek r1 all failed to even transcribe or solve (when given correct transcription). but o3-mini can solve!

[Binnie @AI_Binnie]

@aiDotEngineer @OpenAI @AnthropicAI @pydantic @vercel @neo4j @awscloud @solana @Letta_AI @Google 😭

[Binnie @AI_Binnie]

With Anthropic’s release this passed somewhat under the radar. But diffusion in just 2 steps is crazy and to quote the post it is “opening up possibilities for real-time generation” 🤯

[OpenAI @OpenAI]

Introducing sCMs: our latest consistency models with a simplified formulation, improved training stability, and scalability.

sCMs generate samples comparable to leading diffusion models but require only two sampling steps. https://t.co/rHHSE95sjo

[Binnie @AI_Binnie]

Anthropic’s Computer Use is a sign of the future. It maybe slow, buggy and nerf’d for safety, but I’d say we will have reliable virtual workers, that use apps like humans within a year. It’s going to get wild

[Binnie @AI_Binnie]

@Scott__Myles

[Johannes Stelzer @j_stelzer]

New way to navigate latent space. It preservers the underlying image structure and feels a bit like a powerful style-transfer that can be applied to anything. The trick is to...

[Binnie @AI_Binnie]

Wow, whole song Ai generated from a single prompt. I'd say someone will AI generate No 1 hit by the end of the year. ElevenLabs seems next level, Udio is good, and now has inpainting but this is wild. Here's another one https://t.co/VtKBXpZkDL. How long before AI takes over media

[ElevenLabs @ElevenLabs]

Title: My Love

Style: “Indie Rock with 90s influences, featuring a combination of clean and distorted guitars, driving drum beats, and a prominent bassline, with a moderate tempo around 120 BPM, and a mix of introspective and uplifting moods, evoking a sense of nostalgia and hope.”

[Binnie @AI_Binnie]

@mattshumer_ It is indeed an older project, it's still using openai api pre v1, but that was an easy fix. Loving The workflow though :)

[Binnie @AI_Binnie]

Is anyone going to live in the "real world" in the future? people will live in the overlayed world they want to live in.

[Wenbq @wenbq_me]

This is on Vision Pro🤯

I can't believe I'm seeing THE MATRIX

[Binnie @AI_Binnie]

Microsoft paper "The Era of 1-bit LLMs"  showing no need for floating point weights. 1, 0 & -1 is all you need, Matrix multiplications become additions and it matches Llama performance but 3x faster/smaller🤯 https://t.co/RA5e22fz10. Looks like Phi-3 will be real good.

[Binnie @AI_Binnie]

This is awesome bringing memories to life. Imagine doing it on a movie. Then there is the Genie paper (https://t.co/56HoaiGmLO) turning images into playable games, imagine playable 3D memories or movies. Holodeck is almost here.

[Blaine Brown @blizaine]

This is insane!🔥 This is a 720p (2D) video I shot of my kids 15 years ago! I used AI to upscale it, depth-map it, & used the depth map to convert it to 3D SBS, & then converted that to Spatial Video for Apple Vision Pro (MV-HEVC).

My son is currently a senior in high school, and my daughter is a sophomore in college. This video has complete spatial depth when viewed in the Apple Vision Pro. Obviously, you can't appreciate that via this post. But it's like I'm there & I can see into this room. Talk about giving you the feels.🥹 This can be done with ANY video. 🤯

I'm working on refining the workflow, as it's a bit cumbersome. I plan to share it, but holy cow! I had to post this.

[Binnie @AI_Binnie]

Gemini 1.5 and Mistral Next may both better than gpt-4 and Google also releasing open source Gemma model that beats base Llama 2 and Mistral 7b. I imagine Gpt-5 and Llama 3 won't be far off now.

[Binnie @AI_Binnie]

we spent $1 on ada today so not going for cost saving option😆

[Binnie @AI_Binnie]

New embeddings models save x6 cost for more accuracy swapping ada-002 to 3-small... or more interestingly embedding can be shortened🤯so go 3-large shortened to 256 and outperform ada-002 and save x6 space (and some processing time)🤔or moneybags it and go full 3-large accuracy

[OpenAI @OpenAI]

Expanding the platform for @OpenAIDevs: new generation of embedding models, updated GPT-4 Turbo, and lower pricing on GPT-3.5 Turbo. https://t.co/7wzCLwB1ax

[Binnie @AI_Binnie]

Nice article and collection of resources on techniques to evaluate LLM responses https://t.co/ABezmOgcRf