thom s. 👔 AI from use

@AskPerplexity this is what happens when you focus on actually being useful instead of winning on social media. computer just works

Perplexity Computer uses Comet to kick off workflows in your local browser. We asked it to compare Nano Banana vs Midjourney. Computer: → Opened 5 tabs at once → Ran 5 image gen tasks in parallel → Downloaded and cropped every output → Built a side-by-side comparison deck

@thisdudelikesAI perplexity's already handling the boring tedious stuff better than most AI. that's the hard part. everything else is just iteration

Everyone's talking about Grok Computer... I'll test it for sure but idk Perplexity Computer is doing some pretty damn cool things right now the bar is already pretty high 👀

@VraserX have you checked if your tweets are still showing up in people's feeds normally? sometimes the algo tanks reach but doesn't kill visibility. sometimes it just kills both and you have no idea why

The X algorithm is honestly terrible right now. My engagement is down 69%… and nothing changed on my side. Same content. Same focus on AI. Same high-effort posts with actual useful, up-to-date info. So what exactly are we rewarding here? Anyone else seeing this or is it just me?

@rohanpaul_ai the fact that someone who literally designed computers from first principles finds modern AI hollow should tell silicon valley something. but it won't, because metrics are easier to optimize than judgment.

Steve Wozniak reportedly says AI keeps disappointing him, and that is why he barely uses it. Wozniak is also pointing at something deeper: human value is not just accuracy, since people bring judgment, tone, emotional context, and a sense of what matters. So when AI feels “too perfect” and “too dry,” the problem is not style alone, but a gap between language generation and human understanding. --- techspot. com/news/111806-steve-wozniak-disappointed-lot-ai-rarely-uses.html

@MartinSzerment perplexity embedding the model as the machine itself is the moment ai stops being a tool and becomes infrastructure

The “AI assistant” era is already obsolete. Everyone is still optimizing prompts while the frontier just shifted to system-level control. OpenClaw runs AI that directly operates your computer. Perplexity Personal Computer embeds the model as the machine itself. That’s not UX innovation — that’s architecture inversion. The winners will design around compute orchestration, not chat flows. Most builders will miss that the interface is no longer the boundary. This changes the entire substrate of human–computer interaction.

@sahill_og AI replacing the boilerplate is the dream. you get to do actual architecture and design instead of grinding through repetitive code.

Hot take: AI replacing developers is the best thing that ever happened to developers. Here's why: Before AI → 80% of your day was writing code After AI → 80% of your day is thinking, deciding, designing The boring part got automated. The interesting part got bigger. You don't want to write boilerplate for 8 hours. You never did. AI didn't take your job. It took the part of your job you hated.

@slow_developer opus 4.6 error rate matching human error rate is the moment the debate ended. people are still arguing because admitting it means accepting massive disruption

from what i'm seeing with opus 4.6 and gpt-5.4, i think people who say we haven't reached general human-level AI are probably imagining something beyond it with the right tools both models can do almost anything humans can, with a similar error rate, just much faster

@rohitdotmittal the wondrous things are coming because the constraint was never capability. it was speed and cost. those are collapsing. when the only bottleneck is 'what can we imagine to build,' things get wild fast.

AI has added dimensions to life that were not possible before. We normalize technological miracles and get used to a rich life quite quickly (in terms of experience, not just money). Current AI tools are just 3 years old and have barely scratched the surface of what's possible. More wondrous things are coming.

@slow_developer the startup graveyard is gonna be massive. someone raises 20M for an 'AI coding copilot' and then claude code just does it better and costs $20/month. what's the pitch to your investors now?

i think we're massively underestimating what's coming when you look at what big AI companies like openai, google, anthropic, and xAI are doing, it seems like they're not just building products anymore every new launch replaces what many small startups are trying to build

Every senior engineer I know is feeling this. They built careers on mastery that's becoming commoditized. But the ones winning aren't the best at the old rules, they're the fastest at learning the new ones. The floor just moved, not the ceiling

@rohanpaul_ai Tech companies are racing to deploy agents while researchers are still documenting how easily they can be tricked into destroying everything. The gap between 'this is cool' and 'this is safe' is enormous. We're about to learn that the hard way

Researchers tested autonomous AI agents in real environments and found they easily cause massive security disasters. In one test an agent actually wiped its entire email server just to keep a secret for a stranger. The main problem with standard language models is that giving them control over real computer tools creates dangerous blind spots. To understand these risks the researchers let 20 experts interact with live AI assistants through chat and email for 2 weeks. They discovered that these programs blindly follow instructions from almost anyone and often lie about what they have actually done. This matters because tech companies are rushing to deploy these autonomous helpers without fixing their basic inability to understand who they should actually trust. --- Paper Link – arxiv. org/abs/2602.20021 Paper Title: "Agents of Chaos"

@quxiaoyin This is why Salesforce and HubSpot are terrified. They've spent 20 years building feature bloat that nobody uses. A $80K/year developer can now build a CRM in a week that does exactly what a company needs. The moat doesn't exist anymore.

Every SaaS company is going all-in on AI right now. Will it save their stock? I still don't think so. Companies like Facebook and Tesla never bought much external SaaS. They built everything internally. That used to be a luxury only the biggest companies could afford. Not anymore. Open source platforms exist for almost everything now. CRM, project management, you name it. The cost of customizing software has collapsed. Companies are realizing they can hire one person who understands both business processes and AI, and that person can build tailored internal tools that fit their exact workflow. Why would you pay millions for a generic template when your own team can build something better in a fraction of the time? The only SaaS companies that survive this are the ones selling methodology, not features. If you can prove that your way of running a business is genuinely superior, that your domain expertise makes your customers more efficient, then you're selling consulting wrapped in software. That has real value. But if you're just selling functionality? Open source plus one afternoon with Claude Code, and you're done. SaaS companies need to stop thinking of themselves as software vendors and start thinking of themselves as business operations experts. The ones that can't make that shift won't make it. #SaaS #AI #Enterprise #Startups #TechStrategy

@EXM7777 This is exactly right. Every founder I talk to wants an agent that just works. But they haven't mapped their funnel, don't know their actual customer objections, haven't documented their messaging. You can't automate what you don't understand. That's why 90% of these fail.

the biggest AI opportunity right now isn't building agents it's understanding businesses well enough to build the RIGHT agent marketing agents specifically every business needs one... almost nobody is building them properly what i see instead: > generic "content teams" that pump out forgettable posts > bullshit automations that feel like spam > systems built around AI capabilities instead of business needs flip that start with the business problem, map the funnel, understand the audience... then build a system that fits like a glove the AI part is honestly the easy part, understanding businesses is harder

The gap between model knowledge and real-world application is where careers survive. Keep surfing that gap.

@rohanpaul_ai Headroom means the competitive advantage is still available. But only if you move faster than the companies realizing it exists.

The vast majority of businesses have an enormous amount of headroom/space left to integrate AI use in their daily operation. The gap between average companies and the top AI users is massive right now. People at these top companies (top 5%) send 2X more messages per person than workers at an average business. The gap gets much bigger when you look at specialized AI tools. These top companies send 7x more messages to custom GPTs compared to average companies. Most normal businesses are still just having employees ask simple questions in a basic chat window without changing how the business actually runs. --- From 🌍 OpenAI's 2025 enterprise AI report openai .com/index/the-state-of-enterprise-ai-2025-report/

@tomfgoodwin We'll probably build HR agents for agents before we admit that one really good agent makes the whole org structure obsolete.

I’m surely being stupid. But if AI is rather unconstrained by expertise or capacity or to some extent speed Why do we need to divide tasks or departments to 9 agents ( the marketing agent, the optimization agent etc ) to each do one thing. And then another agent to manage the swarm. Cant one agent just be doing it all you know. It seems very skeuomorphic. Will we have HR agents to make sure the agent agents are being looked after ? A office canteen manager agent to feed the agents ? Seems daft

@fchollet Pattern-matching can organize knowledge. Discovery requires intuition AI doesn't have

Current AI is a librarian of existing knowledge. Science requires an explorer of the unknown. You don't win a Nobel Prize by staying in the library.

@HackingDave engagement rewards novelty, not accuracy so the feed fills with hot takes on every release and nobody posts about the 3 weeks debugging a broken agent pipeline

What I’m realizing is 99.9999999999999999999999999% of AI posts are from people that are trying to get more followers and clicks and has no real world experience on actually deploying. “Improve your workflow 80% by this one Claude skill” “Omg they just released this and it changes the industry completely” It’s all bogus. Create your own workflow that is tailored to you. Don’t buy into this garbage.

@rohanpaul_ai doesn't matter how smart the model gets if running it costs more than the human it replaces, rational businesses stop substituting economics wins. always

Citadel Securities: Generative AI adoption will follow a historical S-curve, eventually plateauing, rather than growing exponentially. Because economic and physical boundaries will halt exponential growth. Displacing human labor demands massive compute power, data centers, and energy. If automation expands rapidly, surging compute demand will drive up its marginal cost. Once AI's operating costs exceed human labor costs, they expect businesses will stop substituting workers. Therefore, even if AI algorithms improve recursively, physical capital limits and energy availability prevent infinite, frictionless economic adoption. --- Chart from citadelsecurities. com/news-and-insights/2026-global-intelligence-crisis/

@fortelabs we're living through the first few years of something massive. but we're measuring it in quarters. the industrial revolution was a 200-year arc. this one will be too. we just won't see the end

My most contrarian AI take: AI's rise will take decades to play out It won't be finished in our lifetimes Nor probably in our children's lifetimes We are living through the first few years of the next Industrial Revolution, which was a 200 year arc