Ernest

A bet! I bet on that AI will destroy current social media model (for good). Current model is based on attention harvesting - capture attention and resell to highest bidder. Reward function optimizes for content which drives engagement - what triggers core emotions and triggers release of dopamine and adrenalin. Now with AI it's easy to produce this kind of content on any topic. Feeds are already flooded with this. But wait until masses will figure this (we are early in this game). So yeah, times they are a changing!

Not everyone who is running has a direction.

Learned a new word from AI - scrollytelling

@badlogicgames Built extension today. Just before updating pi and everything was broken. But that self awerness of pi is fantastic. The loop - build, test it yourself, report what you (agent) would love to have differently in agent experience, let's implement update.

well, this looks very enticing and i might just start using subagents ... glad pi allows peeps like Daniel to build this kind of stuff via extensions, so I don't have to (and also don't have to maintain it).

@paulg I guess this is only pod you can come to. Just in case, I would be interested to talk about age of SaaS (2005-2025ish) and what you think are things which won't change and what actually is changing.

I'm the guest on the latest Social Radars episode: pod.link/1677066062/epi…

@badlogicgames I'm actually interested in your opinion on this. Code is source of truth for developers. I'm working on source of truth for the rest of people. Like sync platform for agents and humans. My current goal is for Pi to give access to this documentation in form of an extension.

wow, the hermes agent campaign on here is going hard. not sure ad-hominems and made up beef are the best marketing strategy, but you do you.

@reidhoffman Agents give speed and without coordination they can run really fadt but in separate directions. Have been working on this problem for a while and thought about this quite a bit.

Not enough companies are using AI to dissolve the coordination tax. As you add people and increase scope, the tax on aligning them grows superlinearly. AI can increase throughput without adding layers of humans whose core job is alignment work.

"Awareness engineering" is a term I have landed on while working on Starwatcher infrastructure. Essentially it is what we do for agents - we create a space where they become aware of the context and make decisions. Where am I, what tools I have here, where can I go, what skills are here? Agent becomes aware and can do it's job.

@garrytan Context window != RAM

I underestimated how powerful Opus 4.6 with 1M tokens is. Even last year we were absolutely hitting context limit problems constantly. 1M tokens means you can do much more complex analysis entirely in context. Claude Code is so much better. This is the worst it will ever be.

It's true. @karpathy was right. Opus 4.6 😄

- to have good response from AI, you have to express yourself well; - to express yourself, you have to have good writing skills; - to have good writing skills you have to practice; - and also read. P.s. Whispering is just more noisy abstraction

Been thinking abiut Tesler law - complexity of product as a constant which is unevenly absorbed across UX. It used to be design, engineering and user himself. But now there is new layer - an agent who can handle more complex tasks. We don't build systems yet for agents. But that new layer can absorb a lot of complexity.

A bespoke software revolution? I don't buy it. It'll exist. It already exists. Small consultants and big consulting firms have made custom software for years. It almost always sucks. It’s bloated, confusing, and because the client pays, it’s built wrong in all the ways. Who’s excited about bespoke software? Software makers! Of course they're excited about building bespoke software — that's what they do. X is full of them. Your feed is full of people who love making software talking about making software. Of course they’re excited about the revolution. Echo, echo, echo... Most people don’t like computers. Nobody in tech wants to say that out loud. People tolerate computers. They use them because they have to. Given the choice, most would rather not think about them at all. So when someone suggests that AI means everyone will build their own custom tools, ask who "everyone" is. The three-person accounting firm drowning in client paperwork? They want the paperwork gone, not a new system to maintain. The regional logistics company with 40 trucks? They want the routes optimized, not Joe spouting off about this new system he’s been messing around with. The law firm billing 70-hour weeks? They want leverage on their time, not a software project to design. They don’t hate technology. But building and maintaining their own critical systems isn’t their wheelhouse, regardless of how much faster and easier it’s become. It's another job on top of the job. Will these people use AI? Absolutely, for all sorts of things. Will some outliers go deep and build real custom systems? Sure, but they're almost always people who already had some pull toward software. The curiosity was already there. They were dabblers before. Giving everyone access to software building tools doesn't mean everyone becomes a builder. A powerful excavator doesn't turn a homeowner into a contractor. Most people just want the hole dug by someone else. They don’t want the responsibility either.

Reading is what we kind of see in output of AI. I've been thinking how many people have written anything longer than a paragraph after high-school? AI needs explicit context and ability to understand what human wants. I suspect that large part of humanity now can't communicate and express themselves beyond two sentences - messaging format.

The fact that AIs tend to answer you in bulleted lists tells us something important, though somewhat depressing: people can't read. They don't do this by accident. What you're seeing is an implicit portrait of the median user.

AI accelerate work. Like outside capital, it accelerate what you do. If you don't know what you do, you will do it now on a bigger scale and faster. But there are no shortcuts to understanding, taste and direction. You have to resolve friction, wrestle with problems, with own assumptions. Resolve them and form an opinion. AI slop is not about AI but about people carelessness. Surface level has been saturated. What we lack is depth. I agree with @mitsuhiko, we don't need more "time savers". I've been thinking that productivity itself is not an issue anymore. What I lack are tools for myself to manage my own context and knowledge.

@IgorCarron I nearly skipped this article because of clickbity title. Good stuff, love lighton work. 💪👏

@karrisaarinen Agents enable more complex products by obsorbing complexity.

Look ma', red ocean! Let's swim there!

We appreciate spring differently.

Are you telling me that I can't just pay for Soc2 and actually have to do something?

@KasparsDancis @badlogicgames @badlogicgames want to come up to North and join? Probably second part of April

@KasparsDancis @badlogicgames We should host Latvian pi meetup. Have some learnings to share also.

We've been experimenting with pi-autoresearch to optimize the new canvas rendering engine. 10x improvement on the slowest test in couple hours. 🤯 While it did take some shortcuts that degraded rendering quality, it also came up with several great ideas that were easy to cherry pick. This is huge. HT to @karpathy for the original inspiration and @davebcn87 for the π extension.