Franktheftank

@franktheftank

97 Following

3 Followers

Franktheftank pfp

We tried building agentic workflow with MCPs, realized MCP only produces an output for every input and then it just shuts down. So it’s very hard for autonomous agent, but good for chatbit

0 reply

0 recast

0 reaction

Franktheftank pfp

Can I get a like?

0 reply

0 recast

0 reaction

Franktheftank pfp

0 reply

0 recast

0 reaction

Franktheftank pfp

0 reply

0 recast

0 reaction

Franktheftank pfp

0 reply

0 recast

0 reaction

Franktheftank pfp

We all won the testnet airdrops

0 reply

0 recast

0 reaction

Franktheftank pfp

Because you got some love from Monad through Farcaster

0 reply

0 recast

0 reaction

Franktheftank pfp

Smoking white through the chimney hole!

0 reply

0 recast

0 reaction

Franktheftank pfp

6/ Building on Search-R1's success, I believe Tool-R1 could be a game-changer in how AI interacts with the world. No longer do we need MCP. Could this be the key to more capable, more natural AI agents?

0 reply

0 recast

0 reaction

Franktheftank pfp

5/ The magic happens in how tools become part of the thinking process. Not as separate steps, but as a fluid part of solving problems. Imagine an AI that grabs a calculator mid-thought, just like you would!

1 reply

0 recast

0 reaction

Franktheftank pfp

4/ Think about how you use tools - you don't follow a manual, you just grab what you need when you need it. That's exactly what we should teach AI to do! Tool-R1 makes tool use feel as natural as thinking.

1 reply

0 recast

0 reaction

Franktheftank pfp

3/ The big innovation? Instead of hard-coding how AI should use tools (which is what most systems do), Tool-R1 lets AI figure it out through experience - just like humans do! No extra steps, no complex rules - just pure learning from outcomes.

1 reply

0 recast

0 reaction

Franktheftank pfp

2/ Here's the cool part: Search-R1 showed us that AI can learn to search the web naturally through trial and error (reinforcement learning). We thought - why stop at search? Tool-R1 extends this idea to ANY tool - calculators, APIs, you name it!

1 reply

0 recast

0 reaction

Franktheftank pfp

1/ Search-R1 is an awesome paper about teaching AI to use search engines through reinforcement learning, we've been thinking: what if we applied this to all tools? Meet Tool-R1, our take on making AI tool use more natural!

1 reply

0 recast

0 reaction

Franktheftank pfp

Had an interesting conversation with colleague who built the earliest visual operator for Google Pixel phone testing. We always thought MCP is overrated for AI, but cannot think of a better way to improve user experience, but a recent paper gives us an idea that could be better than MCP, here is the idea 🧵

1 reply

0 recast

0 reaction

Franktheftank pfp

I like to stay in youth hostels because I never know who I am going to meet there. I made friends who would become prolific researchers and builders in their respective fields that I wouldn’t meet any other ways. Now my crew were mostly hostel frens I made in the past 10 years🏔️🏝️🏜️🏖️

0 reply

0 recast

0 reaction

Franktheftank pfp

Then L3 is very good for local models

0 reply

0 recast

0 reaction

Franktheftank pfp

@carol growth hacking rhapsody with an OG from OG lol

0 reply

1 recast

1 reaction

Franktheftank pfp

Yes, but innovators can leverage that as well. I don’t like nonfungible becomes irrelevant or superficial but hey

0 reply

0 recast

0 reaction

Franktheftank pfp

I think it has a better shot, because it can get the liquidity from cex, and not necessarily be NFT, the NFT is merely a skin, it is in nature a meme project. Non-fungible is a meme in this case

0 reply

0 recast

0 reaction