Astonished that using @GeminiApp to quickly ask questions about the YouTube video I'm watching has roughly 2x'ed my learning rate. Gemini's privileged integration w/ Google's apps seems like a massive competitive advantage on Android.
@OpenAI Interacting with LLMs that use RAG by voice can feel unnatural due to latency. It might be more productive for the LLM to solicit additional context based on prompt while concurrently performing the search. Might feel more natural and be more productive.
I know it's the big cliché to complain about Apple hyping up innovations that Android has had for years, but... I've been enjoying an OLED tablet that's only 5.5mm thick for a few years now. Can confirm reduced thickness and weight are a big deal - worth the extra $
@metaai With all well-deserved buzz for LLAMA 3, I feel like people are sleeping on the quality of your new image generator! Ability to follow complex prompts frequently outclasses DALL-E. No more endless searching for that perfect clip art for a presentation!
@Google_Gemini_ why is there still no calendar access in the GSuite integration? Why can't I create filters using natural language in the @gmail app yet? Slow uptake of Al features in G ecosystem has me eyeing MS despite high switching costs. :-(
@patrickjs@TC39 JS is memory safe. I'm advocating for a broader interpretation of language safety. Would need a type system which can assert stronger control of information flow in-process - similar to how Rust prevents data races. TS could help with new features.
WH now pushing devs to use safe languages. As software continues to eat the world, devs need to acknowledge we have a broader obligation to take security and privacy seriously. New privacy laws mean we need to expand notion of lang. "safety" beyond memory and concurrency safety.
Got an email about a class action against @Shudder for VPPA violation. Can't single them out. So many privacy laws now exist, and yet industry has invested so little in PLs/tools to enable compliance. Safe to assume most corp. software you use is non-compliant. Reckoning coming.
Developing new successful programming languages is almost impossible ATM, because initially there won't be enough training data to enable LLM-assisted coding. This bootstrap problem means PLs need to be 100x (not 10x) better to attract broad adoption. Good luck :-(.
Code-first #graphql servers simplify your life by leveraging your PL's native features to express GQL concepts. Code-first #graphql *clients* can do the same, provided your PL is expressive enough. Example: Zeus expresses GQL queries as JS exps + TS types https://t.co/AGNbr7P9gq
Excited about the JS AsyncContext proposal which introduces an analog for threadlocal in JavaScript's event loop-based concurrency model. While I was at Meta we had some big use cases which would've benefited from this proposal. @littledan
Ahhhh friends, I’m going through a personal renaissance on multiple levels. And for the renaissance that’s most relevant here….my God, what a time to be writing software for the web. Finally caught up on this talk, & holy smokes…
https://t.co/L4YRQHCNMC
Recently rented two bikes with day passes from @BCycle in Austin. I was surprised to get an email saying one of the bikes was not returned! Their docking stations are pretty unreliable. Been getting the run around on the phone for an hour now.
I built a #GeminiAI-backed NotebookLM for the fantastic @RationalRemind finance podcast, and it has been a very useful companion resource while listening. Also great for quickly citing scientific studies when friends/family/advisors propose "creative" investment strategies.
On sabbatical been playing with available AI apps to get a sense of their potential impact on non-technical users today. Takeaway: Post-Gemini, NotebookLM is a good *free* alternative to custom GPTs. Big UI bugs that cause content to disappear during scroll may hurt usage though.