JPEG

Imagine that...

NoBrainer || ||

[JPEG @JPEGx0]

@avrldotdev Before a large task, write tests that are run to verify completion. Once it thinks it's done, it must run the test and report back before final complete is claimed.

[JPEG @JPEGx0]

@Im_IrushiK They should train prediction to auto-route people to sonnet when asking simple things , and give power users random usage bumps. This would keep people happy. But would it break consumer law?

[JPEG @JPEGx0]

@claudeai Remember when Deepseek showed reasoning tokens, and OpenAi o1 was forced to show reasoning cause it wasn't fair to hide it.

Competition is great.

[JPEG @JPEGx0]

@AlexFinn This is the moment.

[JPEG @JPEGx0]

@AlexFinn Remove the tedium from people's lives, built upon the technology stack we've integrated over 20 years. Unstack people and make them pay a tiny bit for it

[JPEG @JPEGx0]

@Lukealexxander This is cool!

Feeding json schemas in/out of models = agents, if packaged up correctly, be used like chainable APIs.

[JPEG @JPEGx0]

@elonmusk

[JPEG @JPEGx0]

@elonmusk @realDonaldTrump

[JPEG @JPEGx0]

@elonmusk @realDonaldTrump

[JPEG @JPEGx0]

@gregisenberg @meetLCA If you're making product now.
AX (Agent Experience) matters also. How does AI navigate and understand your product... for a user

[JPEG @JPEGx0]

@fvd_explore @DaveShapi @kimmonismus But the phones and computers are already in our structures. Now it's just machines operating them vs ourselves. I think this changes adoption timelines

[JPEG @JPEGx0]

@elonmusk This is the beginning of Corpocracy.
The company is more capable than the government.

Amazon took a global marketplace to your door.
Meanwhile Tony & George took us to war.

Google all the knowledge in your pocket
Politicians keep trying to stop it

[Birchlabs @Birchlabs]

Meta have shared an LLM pretraining codebase
https://t.co/2mxq3IDHd1

[JPEG @JPEGx0]

@7etsuo @icodeagents "my AI occasionally does this git comet thing. then when it fucks up, says git push or pull or...something"

[w̸͕͂͂a̷͔̗͐t̴̙͗e̵̬̔̕r̴̰̓̊m̵͙͖̓̽a̵̢̗̓͒r̸̲̽ķ̷͔́͝ @anthrupad]

It takes a flipping-over of perspective when dealing with complex systems - switch from averages to multistable basins, niches, regimes, etc

[IntellectualDorkWeb @Intel_Dork]

😂😂

[kelin @kelin_online]

teaching design students to code with LLMs, 1 month update:

1. projects are a lot more ambitious than i expected

they don't know what's hard, and AI will tell them to use a certain tech stack, so they just figure it out

[JPEG @JPEGx0]

@icodeagents Card Counting - A method for dynamic persona system prompt self adjustment.

Fine-Tuning unique tokens for personality quirks... inspired by Entropy Based Sampling

@Orwelian84 and I thinking about it

[JPEG @JPEGx0]

@sdrzn Was really waiting for this. Thanks!

[JPEG @JPEGx0]

@sdrzn Just wanted to say a massive thank you. Your work is exactly what I've been looking for!