Vlad Bars

The gap between the top two and everything else is becoming the only conversation that matters for production AI. When you're building products where output quality directly impacts user trust - healthcare, fintech, legal - "good enough" models aren't good enough. The choice is between these two or accepting a quality ceiling your users will feel

My big takeaway is that both Sol & Fable represent jumps over previous models and have opened a large gap with the next-best AIs. People will have preferences for one or the other, but if you doing any work where better intelligence matters, those two models are your only choices

@DotCSV Nice view!

Aún por Kenia! Tranquilito el mundo de la IA no?

@strimblez Makes sense. When models are commoditizing every few weeks, the only stable differentiator is the relationship. Same pattern we saw with cloud providers ten years ago - AWS vs GCP wasn't about specs, it was about whose solutions architect answered your Slack at midnight

multiple founders over the last few days have told me that their decision to build with OpenAI vs Anthropic is increasingly not technical, especially as the models seemingly “leapfrog each other every few weeks.” rather they say it’s about the experience theyve had with the teams at each lab - how much attention they’ve been given, how they’ve been treated, and most importantly, the vibes. thoughts? 👀

@iamtrask isn't the semantic structure in gradients just a necessary artifact of optimization, not evidence of anything resembling consciousness? A compression algorithm has structure too - does that make it conscious?

why is an LLM storing information in first derivatives related to consciousness? LLMs literally cannot learn anything that wasn’t first encoded as a gradient. so yeah… the gradients will have semantic structure… otherwise the LLM literally couldn’t learn anything. thats backprop.

@svpino Curious about point 4 - when learning spreads per-team vs per-user, how do you handle the moment one user's correction contradicts another user's preference within the same team? Who wins and how does the agent decide?

How you can build a moat with self-learning agents: If you can build an agent that gets better every time people use it, you will be unstoppable. Here is what you can do: 1. Learn from two sources, not one. Agent traces show what the agent did and where it broke. In-browser activity shows how users steered and fixed the results. Most products capture the first source and forget the second. 2. You have three options to apply new learnings: fine-tune your model, update the harness, or provide in-context information to the agent. Make sure you take advantage of all three. 3. When deciding what your agent should remember, lean on procedural memory (where you store workflows and rules) and episodic memory (where you store specific things that happened). Don't over-index on semantic memory, as it can go stale and become a liability. 4. Set boundaries around how far learning spreads (per-user, per-team, per-app). Be careful not to leak facts between different users' contexts. 5. You want to own the data your agent learns from. This data is your most valuable asset. Try to keep it in your infrastructure so it doesn't get stolen by Big Cloud. 6. Use open, framework-agnostic standards to capture and apply learnings. Here is an article with more information:

@tunguz The moment AI starts optimizing interfaces for other AI agents instead of humans, UX design becomes prompt design and the entire frontend layer becomes optional. We're closer to this than most people realize

You guys still don't get it: before this year is over, the best software will be made by AI *FOR* the AI.

@anulagarwal Сongrats!

new mobile app submitted for review! hoping it passes lol

@Tito_design_ Like the app and UI.

Mobile App - Pilates

@hthieblot Build the app in an hour, get rejected for "guideline 4.3 — this app is too similar to apps you haven't built yet" 😄

AI takes 1 hour to build the mobile app Apple takes 3 weeks to review it

@rudykadoch @InsilicoTrading @MetaMask Matching pro terminal revenue per user at half the fees - that's a counterintuitive result most fintech founders wouldn't expect. What surprised you most about the user behavior that drove that? And going global which market do you think will respond first, and why?

A year ago, I would’ve never believed a retail mobile app could match the revenue per user of pro trading terminals (@InsilicoTrading) while charging half the fees of @MetaMask. Yet here we are 🎉 Last year was the warm-up. Now that we’ve broken every barrier holding us back from reaching a much larger market (chains, crypto-only UX, gas tokens, etc.), it’s time to go global! @massdotmoney

@JameslabiQ @marcelkargul Nice UI

Hands down, the best mobile app we’ve ever built. By @marcelkargul

Productized service. No build time, no App Store review, no infrastructure. Define one specific outcome for one specific customer type, charge for it, deliver it once. Learn what people actually value before writing a line of code. The app comes after you validate the problem is real

If you had 30 days to make your first $1K online, what would you build? 1. Mobile app 2. Web app 3. Productized service 4. Content/tool hybrid I’m choosing one and documenting everything publicly.

@ilyosme Help founders to build apps and scale them to millions of users. Let's connect!

I want to connect with founders If you are building mobile app or web app, lets connect If you have friends who are building tag them too

@rj_26_10 Nice! Cheers 🍻

GM Off to a slow day today! More pace expected in the later part of the week. :) Cheers! ❤️

@conshuz Let's smash it! 👊

Gm from your friendly neighborhood ☀️ Happy Wednesday, let's smash it 👊

@JBYmetaX GM!

Morning fam ❤️

@aira_tubby Active

Below 9k followers ❕ Just say "Active" We'll f0llow you

So, the initial findings are: - Add-to-cart rate: 11% on desktop and 5% on mobile. - Conversion: +55% from the 1st iteration. Built sophisticated and custom analytics specifically for this page. AI A/B tests, bounce rates, micro-engagements & hesitation, bot traffic, rage/dead/quick clicks, bounce rate for each section, healthmaps, perf analysis, and UX analysis, CTA comparison, and so on. Everything is AI-analysed. Preparing the 2nd iteration in a couple of days. Stay tuned.

The healthcare business sends traffic to a landing page at more than 8 figures per year. They said conversion is very low (1-2%) They wanted to 2x or 3x it. Thought UX, UI, and page load don't matter. Until now. We audited the page across more than 80 checkpoints. Rebuilt it from scratch. Improved UX, UI, and load speed. Now it's been two days. Findings and client feedback in the comments.

@mishi_not Good evening😀

Good Evening 🤌 Shaam ki chaye ho jaye ?