Prashant Mittal

Wife: "I need to add subtitles to this video" Every tool: paid, low resolution or watermarked 🙃 So I built one over dinner. ClaudeCode + Whisper + WebCodecs, runs entirely in your browser. Try it → notshant.xyz/autosub Behind the scenes → notshant.xyz/posts/autosub

With all the voice and chat interfaces taking over the app UX, I have been thinking more and more about how a well designed tactile software is actually a work of art.

Can somebody tell me why this literal kick on Bruno in the Man City box, wasn’t checked on VAR or even talked about by pundits? I’ve seen endless replays of Dalots high challenge but the media is silent on this?

We just got featured on @salesforce official customer stories page as an Agentforce implementation partner. Here's how we helped TASC Outsourcing achieve a 6x response rate with AI agents. Read more: salesforce.com/in/customer-st…

What if your next @salesforce project took weeks instead of months? Join our webinar to discover: - How to get working prototypes BEFORE committing budget - Real case study: Agent Force delivered in 1 week - Why AI-first delivery is transforming companies Register now: realfast.ai/exo?utm_source…

@harshh_jainn Same bro

me @ remarkable

@harshh_jainn I was so looking forward to see one of these during my SF trip, but didn't.

incredible

@gonsalves_r @sidbetala sadly, the segment doesn't exist anymore. Closet would be i20 Nline.

@sidbetala Yeah. And not a single good looking hatch. Where’s the Polo of 2025?

So weird. Trying to replace our ~8 year old car; can’t seem to find a single good looking car at a reasonable price point (~20L). Everything is just…weird.

@davidgiller Is there a doc for this or is this not release yet.

Preview Agentforce Headless API | Agentforce Decoded #Salesforce" target="_blank" rel="nofollow noopener">youtube.com/watch?v=4OhbN0…

2024-08-29 edition is out! Read it here: archive.localfirstnews.com/archive/202408…

Join in to see Creole Network in action

Nothing to see, just @prashant_mit dropping Rolodex alpha. Search & discover the right introductions in large, busy communities. Does your Telegram GC need a Rolodex? Sign up: rolodex.social

@AjeyGore @Sinister_Light If you are using tailwind, I would suggest checking out daisyui saves a lot of handcrafting for common components and theming work

@Sinister_Light Yeah, clean bold in this sense, but hering hang of, too much to type, but i think i will be okay in few days, much flexible, but more to handcraft :-)

Tailwind requires tons of classes to remember, i like the extreme flexibility but getting done on front end has become its own chanenge now :-) any good places to get going with tailwind cheat-sheet

@geoffreylitt @Andre_Terron that's why we are trying to build Rolodex (rolodex.social) as the killer app to bring user's social graphs to this shared data layer, built on top of Creole Network (creole.network)

@Andre_Terron Tactically I think the best options are: 1) existing huge platform imposes it. Eg, Apple could do a lot better here (their work on Health has some nice examples of data sharing) 2) new "killer app" - get adoption for an app, which is actually built on a shared data layer

# Ephemeral software with shared data An intriguing possibility: what if people using AI to build personal software tools leads to a next generation data architecture without cloud app silos? A diagram of how that might happen: Breaking down each level: Level 1: "just UI". Starting super simple. We can use AI to generate a UI -- no storage, no backend, no multiplayer. This describes a lot of the incredible Artifacts / Websim examples flying around right now, as well as many of my past experiments building software tools w/ AI. It's straightforward to run and deploy a web app that's "just UI". These tools can be very useful for quick tasks—esp if the UI supports import + export w/ copy-paste or file download. But obviously there are UX limits if you can't save data or collaborate. Level 2: "Siloed apps" If you add data storage + collaboration to the stack for your generated tools, then you can make true "apps" with deeper functionality. This approach seems less common at the moment -- likely because it's more complexity for the AI model to manage, and also more work to deploy / operate. But it seems likely grow to over time. All the existing contenders in the "simpler multiplayer apps" space have something to offer here in terms of cutting down complexity --backend platforms like Replit, no-code tools like Glide and Retool, local-first sync libraries like Automerge...and indeed many of those players are doing very neat stuff. Still, there's a problem: each app is a silo; each app stores its own data. This is totally normal and is what we're used to in cloud apps. But, it cuts off a key use case: making new ephemeral tools for your existing data! (APIs are nice but often aren't powerful enough to truly make new tools) Which brings us to... Level 3: "shared data" Imagine that all our apps were just UIs on top of a shared collaboration and data storage stack -- analogous to the role the filesystem used to play as a shared data layer for all your desktop apps, but reimagined for modern collaboration... Now: imagine you're hitting limits with a tool you're using. What if you could just generate or customize your own alternative tool on the fly, with AI assistance? Perhaps you just add a specialized button that handles one workflow, perhaps you replace the original tool entirely... The key is that you're not starting from scratch with fresh app and no data, you're just tweaking and extending the tools you already had. This shared data layer thing is not a new idea at all, but so far we've remained largely stuck in the equilibrium of cloud app silos. Which makes sense -- it's a hard equilibrium to escape! Privacy / security is (very) tricky for shared data between apps. What's the billing model if not cloud SaaS? And where does this shared data layer live and what can it do? But... I wonder if the energy around AI-generated ephemeral tools will finally be the push that can nudge us towards this different future. Here's the bet: if AI dramatically lowers the cost of UI creation, that incentivizes open environments with shared data, where custom tools can be more valuable. (Sidenote, this is why we pursue malleable software and local-first software together at Ink & Switch, there are strong resonances... will likely share some more demos on this line of reasoning soon 😎) Would be curious to hear reactions / counterarguments to this view!

@harshh_jainn I feel it is a waste of time for both parties involved. Ideally the period should be as close to zero as possible (a week or two at max)

the next 3.5 months are peak shipping season at work and boy are the incentives really bad aligned. interesting how this pans out for all parties involved.

Rolodex is the community director for Farcaster built by @SageiEar. Rolodex allows community members to find the right intros to other members within their community. 🔗: rolodex.social

@adadithya Because that can happen sitting in your home, 1 reel at a time. 😋

The urban millennial prefers wanderlusting over travelling

@alexanderchopan Is this a play before the token release for the network and slowly disengaging into a foundation/dao? A re-org.

@ephemera/gm-ephemera?referrer=0x2f60d2bb84eb8df6951f7215ef035ef052ba2725" target="_blank" rel="nofollow noopener">paragraph.xyz/@ephemera/gm-e… “The name also reinforces that Ephemera's core role in nurturing XMTP is not meant to be permanent.”

@harshh_jainn That's cloud, wait till restarts solve most of the issues and you keep wondering tf happened. enjoy!!! 😂😂

azure is genuinely trash, the same config works some time and crashes the other time. i have literally no idea why my code is working/not working

Portable private social graphs eliminate the need to rebuild your network on every platform. Connect once, connect everywhere.