@Sentdex we don't necessarily need user facing products to solve public good problems - aside from the fact that labs too are businesses. dude's not a technologist, yea he could've avoided 'agentic' but fixating on it misconstrues the core point
@nikillinit folks probs complained about slow typing & network connection & tiny capacity when email first came out. we will know when AI is adopted the moment everyone stops talking about AI
everyone wants technical moat off the bat. i don't think it's a thing anymore mate. in prior tech renaissances (computing, web), github, cloud, and arxiv were not a thing. name one company post 2010 with technical moat that didn't take decades to build
Both Figma and Paper have MCP servers that give a coding agent a suite of design tools and claim to produce better frontends than writing CSS.
We ran an independent eval in Braintrust to test which is better. Here's what we found.
- Paper holds its quality across page types, while Figma drops on the complex designs.
- On both ordinary and design-heavy webpages, the design MCP doesn't matter. Across 27 designs, Paper averages 0.716 and Figma 0.679, and the paired difference does not clear significance (p = 0.21).
- Figma's run-to-run variance is about 1.9 times Paper's, and it runs 42% longer while costing 32% more per point of visual quality ($3.73 vs $2.82).
yes to ai companions that help you learn or fill a relationship no human realistically can. like a toy teaching language to a toddler. or a cognitively dependent senior who needs superhuman attention. to offer this to a healthy adult or senior is laziness and careless intention
@chrissyfarr cynical me is not so sure. with oss models getting good cost will get to minimal, especially compared to human admin. Its possible with LLMs we stay where we started
pretty please release just 1 multi streamed duplex model anthropic or openai. these turn based models cannot consume while they compute and we're hacking our way to robots that can interact to some acceptable social degree