@quxiaoyin next -3 yrs business conversation: applying or RL, -5yrs replacing applications, and robot/edge or everything. will it matter cn or us? labs will be valuable if it is the provider of everything compute. Two forces are after โeverythingโ : model and hardware. It will happen.
@thdxr Yes, there are only 3 models of cars๐ from 2 companies in world. But there are only 2 mobile OS, 1 browser with couple of children. But there are 100s or more CRMs in world with customized for every business.
@BrianRoemmele In one sense, it is free/subsidized consulting to acquire capabilities/data behind successful PE owned portfolios, PE are hedging the risk, labs are acquiring/building model/app.
Today heard that frontier companies will make โsoftware god modelโ that build software for any use case with winning source, and their consulting companies are to get that โsecretโ sources from enterprise and lock ๐ in enterprises.
@twistartups@0xcarro Garage data center was idea, before I moved my current work. Given current price increases and needs of RL env to build winning model, I should do further. At that time, my challenge was how to secure customerโs data. But itโs oceans of synthetic data and (semi) public data
Started using , โโโ triple double quote a lot to let agent knows pasted codes. What would better option? Tested === before but sometime they take that as input.
@svpino 12 agents part: Will it possible soon or possible for project simple enough or tooling enough? I guess it might possible. I am running 9 agents(5 work, 4 wait review) using coordinator agents but it is stressful even review review.md or open new issues/requests.
@SpaceNews_Inc At the same time, I will enjoy when this burst when innovative tech comes and say โthere is new energyโ not nuclear โข๏ธ not oil not โฆ
@SpaceNews_Inc Looks like lot of work coming to space, there were movie, cleaning space junk for $$$ Hope not those policy guys asking upfront env fee for โป๏ธ and โ๐พ innovation.
From yesterday, azure gpt 5.4 api is lazy, every other requests got
Iโm sorry, but I cannot assist with that request.
Tested other versions, same.
Create new deployment, it fixed.
It looks like almost monthly event. Can someone clear logs ;-)
@Azure
Morning emergency with litellm, luckily was one major version behind. Checking dspy, looks ok ๐. When the world of dynamic code, this looks like major risk as it would spread very quickly. if attacks prevent updates?
It is equal or better, when you rely on it to write order with how/skills, you approve it is better than you. You=we
Not mean, we are less, we can do better things.
After reading many arguments what is OpenAI, Google or other LLM providers, they are aol/compuserv of 1990s, and open source models are early tcp/ip of 1990s. Who is coordinating RFC for AI days. We have many rfc candidates, OpenAI API spec, mcpโฆ
@dani_avila7 After that mcp hub/tool calls solved, find issues with common mcp call back and forth, so had to add common context and context compress on top of that.
@farzyness ๐ฏ Human will create and find work and purpose during that transition. Some task moved to AI and robot, like we did outsourcing, some will be created.