zed

From idea to thought experiment to prototype to framework to whitepaper to app store to #hoserstack to possible #tradeirritant ? I am pleased to present Chrrpt 5.0 for your feedback & thank you for your indulgence. #fingerscrossed 🤞🏽

The Federal Government of Canada just gave @Cohere : Canada's version of Palantir $1.6 Billion for I don't what. But don't mind me I am just over here wiring up 7 Different AI Models together into a chat for an actual purpose. 🪩 👐🏽 I have been chrrp'n on here about AI stuff. Local Model this. AI that. Vaguebooking like a mofo. The exact kind of thing I used to rail on if it wasn't me, I'd be annoyed. But it's me, so you know it's going to be awesome. So here it is. Que'est qu'fck? A multi agent AI war room for your Brainstorming & Strategy Work. The video above shows the HAM chat possibilities and is the basis of the Decision Engine for Monolith and only one of the dope features in the kit. Imagine 6 models in one massive multi agent chat feeding into a document pipeline that outputs a doc for you to walk with. In non nerd: The big 3 Frontier Labs (Claude, Chatgpt, Gemini) and 3 separate local models (Byteshape's Canadian Variant of Qwen, Granite and K2) debating that product overview or a client's creative brief that you need done. In a standup that is a whole group people chatting about nothing for most of it. I figured it out. And I just put them in your pocket. (BTW Cohere didn't even break top 20 in my testing). To be clear, I personally have no issue w a corporation doing its thing. This is clearly Corporate Welfare and good on them. Take it from a guy who sold his startup to the feds many moons ago, big league money like this comes with a lot of strings. The biggest regret I ever made was selling my company to the federal government and then having them as a boss was a crash course on exactly why this country has and will never have a serious cultural policy. Origin Story After building the most antiAI product I could (I mean Chrrpt literally lives outside of the Cloud Act so you know I am cooking w gas here) I built a product with ALL THE AI. I was just messing around cause that is what is done now when I discovered LocalLLM stuff and just how good it is. The thing that tied it all together was a Canadian Variant that really delivered by punching waaaay above its weight. Like the tests I ran for Humanities & Knowledge Work stuff which is my area of focus, it almost matched Claude which is my GOAT for game theory and humanities output. Oh, it's a local model that lives on your machine costing you nothing. Read that again. A Canadian Variant is almost as good as Claude in my knowledge work testing. Well done @ByteShape and specifically thank you to @AllenLau for putting this on my radar. What's Next? I am going to be putting these models through a humanities and knowledge work gauntlet over the next few weeks plus these buggers need some etiquette lessons and manners maybe. So after the fine-tuning, I drop Monolith as a fully local and basically Air-gapped AI kit so that your stuff is safe from prying eyes. So my many decades of useless knowledge about Strategy, Advertising and Pop-Culture that was really only useful for Jeopardy after LLMs deaded my career is finally good for something and will not be going to waste. My mom is less worried now.. :) Then Culture. Poptalk is coming back. Spoiler Alert : An arts fandom AI powered by user habits. #giddyup #BuildInPublic

530 is now 430 but how the hell does it feel like 630?

The only kpi that truly is the fistbump you get from the user for solving a pain for them. #thehiphopway

👊🏼

You about to see fandom let itself be known. My prediction is that Netflix buys skydance for the money they were given to walk away.

buffffffaallllllllooooooo 🐃🐃🐃 #gobillsgo

This is clever. Screen sharing can be pretty jank a lot.

not bad @allenlau and @ByteShape and co.. Not bad at all. 👊🏽 #localllm #canadiantech #buildinpublic

RIP Coach 👊🏼

👋🏼