One of our first @vercel taglines circa 2016 was "realtime global deployments".
Global spoke to the eventual convergence of CDN and application hosting, which we now call the frontend cloud.
Realtime is even more exciting… it speaks to shipping software with the speed of FTP'ing a .php file and seeing it live on the next request. Or editing a Google Doc.
We're on the cusp of bringing this dream to reality. I can't wait to show you what this looks like at t.co/MKF1RbJJyF!
Seeing is believing. Something I didn't anticipate is how much @v0 adoption there is in field engineering orgs (solutions, sales eng, pro serv).
It helps the customer *see* what they're buying, and as such it's shortening deal cycles and speeding up migrations.
Ship Weeks are good, but they’re mostly about extending the product surface.
You also need Fix Weeks, with the emphasis on making the existing product great.
The future of the Internet is generative and personalized to the individual.
Previously a lot of human hours went into creating and curating content into static representations. Take t.co/FxTHr2q5CU, an effort to re-write articles into “Simple English”.
Today, you go to an assistant and ask: explain concisely what Hawk Tuah is. You’re very unlike to instead go the route of finding “Simple Wikipedia” in a search result. This is a concrete example of the erosion of the popularity of the search UI in our new world.
What’s even better is that as your history and relationship with the assistant evolves, so will its context about your level of understanding and expertise in the various subject matters. The explanations will have metaphors tailored to you, exclude redundant details, strike the right tone.
Companies like @zep_ai, @mem0ai, and others are building the infrastructure to help ai developers capture memories and preferences to enrich queries and produce even better results. It’s arguable we’re under-utilizing the power of existing models, by limiting them to operating on stateless and static prompts. I’m excited to see the next generation of personalized AI.
Long-running compute should be accompanied by a realtime stream of events to the frontend. This will become even more important as AI models begin to spend more time (minutes, hours, days) working on problems.
The transition from search to AI chatbots might end up having to do less with the AI part, and more with the fact that chat is the most relatable user interface, especially for younger generations.
The vast majority of mobile apps, and the most cumulative user seconds, all involve chat UIs in some way shape or form. iMessage, WhatsApp, Messenger / FB, IG, Telegram. It’s all chat. Embrace the chat.
Very excited about this ongoing compute migration, which debunks some myths about performance across cloud infrastructure providers (colors relevant). Can't wait to ship the blue line for everyone 😁
The fact that @polymarket has taken crypto mainstream in 2024, without people even realizing it’s crypto to boot, is astonishingly under-discussed. We’re so early.
It’s paradoxically become harder than ever to buy online because of all the fake reviews and fake endorsements. Great opportunity for e-commerce focused AI companies to pick up signal from the noise and deliver a quality buyer experience.
All Apple needs to do to sell me a product is to make it in black. Then the next generation needs to be slightly less black, say, Midnight Blue. And then the subsequent one should be all black to bring it home.
4 years ago on this day. CA wildfires, COVID-19, financial turmoil. Good reminder to remain optimistic and long-term focused. Life’s so much better now
We received over 200+ CFP speaker submissions for t.co/bmfw5SUTco – this will be our highest quality conference ever from a content perspective. Thank you all for your submissions!
This is a remarkable idea. Turns every website into an API. Think of it as automatic Structured Output for any website on the internet. Wild. t.co/SRXDuux1ne
Our engineering leadership team is having an offsite in San Francisco. We want to spend time with customers on-site on Sep 26. If you're using or considering @vercel and you'd like us to stop by your office to share feedback, demo to us, or simply roast us, DM or reply.
There's a pernicious edge case with opengraph images… when the image goes over a threshold in size (kb), social card validators will gladly render it, but social networks and chat apps (understandably) won't. @vercel's OG tool will warn accordingly to prevent debugging hell.
