• In case you missed it, LLMs have become cheap. The cost of inference has decreased 100x over the past couple of years (~$50 to $0.50 per 1M tokens). This is the largest deflationary technology trend we’ve seen in our lifetimes (@berman66)
• The next evolution of Claude artifacts, or ai artifacts will be the removal of web from webpage. Simply a page. A new primitive, like a note, document, file. The same way you can write a new note or make a new file you can prompt and interactive page into the existence (@NickADobos)
• LLM + pure code = is the new « no code » — If you are a beginner that want to build app and don’t know how to code I truly convinced that LLM generating code and learn how to interact with code base through tools like Claude is deeply more efficient that learning no code tools like bubble (@BrivaelLp) (24.08.14)
• I bought into the Twitter hype and tried coding with Claude 3.5 Sonnet. Holy shit, I feel like a kid on Christmas morning. Can’t stop dreaming about random things I want to build (@manosaie)
• Computers will control the future… They already do (@elonmusk) (24.07.11)
• There’s a lot of talk about AI-generated influencers “tricking” people…But are they that different from human influencers, who also don’t interact with 99%+ of followers IRL? (@omooretweets) (24.07.12)
• Multimodal fine tuning is going to take so many of the hardest AI use cases and bring them into the realm of possibility (@OfficialLoganK) (24.07.11)
• Many people hold the view that no-code tools mean simplicity and efficiency, while coding means complexity and slow progress. The reality is not black and white. If you have enough experience, you should understand that complexity does not disappear, it just takes different forms. (@JoshGuoDesign) (24.07.11)
• This is what the future of software development will look like. AI writes 80% of the code. Human devs finish the last 20%. The next wave of models will begin to unlock the potential of these tools. Massive AI codegen wave incoming (@mckaywrigley) (24.07.18)
• Building a demo app today develops your intuition and preps you for a time when software using LLMs and tools to execute sequences of complex tasks becomes the backbone of every AI system (@nickfrosst) (24.07.30)
• "Higher intelligence covers more ground in future situation space using the same information.” — LLMs are not intelligent, but they are still amazing because they have the situational context of the entire web. It seems to me that all the advancements we see involve further engineering the context to enhance performance (e.g., RAGs, Chain of Thought) (@sunfanyun) (24.08.04)
• Predictions for the future of software engineering (@russelljkaplan)
• What will happen to Google & OpenAI when open-source catches up? Few Thoughts (@tedx_ai) (24.08.06)
• Notebook for seeing with language models — Investigating how computers could help us not just store the outputs of our thinking, but actively aid in our thinking process (@thesephist) (24.08.07)
• Right now the advantages from AI accrue to workers, not firms. As a result, they can be hard to notice (especially if you don't try to seriously use AI yourself, which is illuminating) (@emollick) (24.08.06)
• How I use “AI” (Nicholas Carlini) (24.08.07)
• I don't know about you, but I'm totally fine with the GenAI bubble bursting. I've gotten everything I needed from it: 1) Apache 2.0/MIT LLMs that I can use to solve business problems I couldn't dream of solving before, and 2) super-fast open-source tools for running these LLMs at a scale of millions of documents per hour. Thank you all for providing these tools free of charge and keeping me excited for two awesome years! (@burkov) (24.08.07)
• When an expert says "an AI can do 80% of the work" on a project, that is not usually a Pareto problem (leaving the hardest 20% to humans), it usually means "AI does a lot of the annoying work and my expertise allows me to spot any errors, fill in the extra gaps and add insights.” (@emollick) (24.08.07)
• Eric Schmidt’s AI prophecy: The next two years will shock you (24.08.17)
• Things like FastHTML + copilot are allowing people with skills on the left to shift right easier than ever before (e.g. Python for everything). Previously I felt like it was too much surface area for mortals to span the entire width of the diagram, but it seems way more tractable now. I suspect there will be tools for people to "shift left" too (but I have far less insight into who is building those) (@HamelHusain, Diagram from @swyx)