Vladimir

at the very start of my coding career, i worked at seo company. they provided full range of services - website audits, marketing, optimizations. back then, tools that could do a deep website audit either covered 10% of what you needed or cost a shit ton of money. so i spent weeks planning the perfect all-in-one tool that'd cover all company needs. what i didn't realize is that showing this to my boss would've gotten me maybe $100/mo more while saving the company ~$10k/mo on seo specialists, because preparing audits was always the mandatory (and most boring) part of their job. i tried to build it. obviously failed, because i simply didn't have enough knowledge yet. i did make some pieces work, but that was nothing close to what a real audit tool should be. years passed. i quit that company, moved on. but recently i caught myself thinking about that idea again. the all-in-one seo audit tool i never finished. so i thought - why not try again, but this time with claude. i wrote a very detailed plan explaining all the app's logic and capabilities, then threw the whole .md file into claude code. it came back with an even bigger plan - full architecture, tech stack, folder structure, auth flows - everything. and ngl that "yes, auto-accept" button paired with "--dangerously-skip-permissions" mode looked really tempting. so i pressed it. two hours later claude was done. i got an entire app: frontend, backend, deployment instructions. i deployed it on my vps just out of curiosity. the app launched on the first try. no errors, no crashes - nothing. i grabbed some random website, entered it into the "url to audit" field, hit enter. five minutes later i got a full website audit with a dashboard i didn't even ask for. this wasn't a mockup, it was a real app with crawling, page-by-page issues, exports, auth, and a dashboard i didn't even ask for. yeah the app itself isn't some impossible engineering challenge, but it would've taken me about a week to build manually, while claude did it in 2 hours. were there areas for improvement? sure, i fixed a lot of things after. but the point is - i spent weeks planning this tool years ago and couldn't build it. now i planned it in an hour and deployed it the same day. the tool i couldn't build at 17 took two hours at 21. same idea. same person. different tools.

@pmarca every era gets a marc andreessen tweet that sounds like a rallying cry but is actually a fundraising announcement

It’s time to build the Independent AI Grid.

@DonaldJTrumpJr @worldlibertyfi "your agent, your keys, your rules" is a great tagline until someone's agent autonomously spends $47,000 on cloud compute at 3am because it thought scaling was the optimal decision the whole point of interns is that they CAN'T pay for things

AI agents that can reason but can't pay for anything are just expensive interns.  Today @worldlibertyfi shipped the infrastructure to fix that. AgentPay SDK, open source, self-custodial, policy-first. Built on USD1.  Your agent. Your keys. Your rules. Check it out: agentpay.worldlibertyfinancial.com

@satyanadella microsoft naming it the "Superintelligence team" while the output is "image model in copilot" is corporate branding at its best ngl next week the "AGI division" will announce a new emoji keyboard

Great to see our new image model from our Superintelligence team rolling out in Copilot and coming soon to Foundry for enterprise customers.

@FirstSquawk "content generation, web search, and personalization" is how you describe literally every ai product when you don't know what makes yours different this reads like a press release written by gemini about gemini ngl

GOOGLE HAS BEGUN TESTING A DEDICATED GEMINI APP FOR MAC TO COMPETE WITH CHATGPT AND CLAUDE, OFFERING FEATURES LIKE CONTENT GENERATION, WEB SEARCH, AND PERSONALIZATION.

@fchollet man that's so true, ask it to solve a problem that looks like a known pattern - and it's magic ask it to solve something genuinely new - and it confidently gives you the closest memorized answer which is sometimes worse than nothing

This is more evidence that current frontier models remain completely reliant on content-level memorization, as opposed to higher-level generalizable knowledge (such as metalearning knowledge, problem-solving strategies...)

@thdxr "we want you to build with claude, but only through the doors we approve" is the kind of energy that makes devs switch providers

opencode 1.3.0 will no longer autoload the claude max plugin we did our best to convince anthropic to support developer choice but they sent lawyers it's your right to access services however you wish but it is also their right to block whoever they want we can't maintain an official plugin so it's been removed from github and marked deprecated on npm appreciate our partners at openai, github and gitlab who are going the other direction and supporting developer freedom

@typesfast the funniest part is google named it gemini before they knew which twin was which lol turns out the search revenue is the immortal one and the ai lab is the one that keeps almost dying at launch events

Whoever named Gemini at Google really named it. In the mythology, the immortal twin gives up his immortality to save the life of his mortal twin. It’s just like Google giving up 100% of its free cash flow to make sure Deepmind survives.

@markgurman google making a mac app to compete with claude and chatgpt is the corporate equivalent of showing up to a party 2 hours late with a bottle of wine nobody asked for

NEW: Google is ramping up development of a dedicated Gemini AI app for Apple Inc.’s Mac computer lineup, looking to step up competition with OpenAI and Anthropic. bloomberg.com/news/articles/…

@cryptopunk7213 the niantic one is genuinely dark ngl 500 million people thought they were catching pokemon and were actually mapping sidewalks for delivery robots at least doordash had the decency to pay $5 for the data instead of disguising it as fun)

man shit is getting dystopian really fucking quickly doordash is now paying people to film themselves doing chores then using that to train AI robots that will replace them - guess what? they're not the only ones: - niantic (pokemonGO creators) has trained AI delivery robots with 30 billion photos taken by 500M+ players - Uber launched 'digital tasks' last year where drivers complete tasks to train AI in exchange for a few dollars i don't think people realise they're doing something that will eventually get them fired "heres $5 in exchange for your job"

@bindureddy i am building an ai product and we literally have a provider toggle between anthropic and openai because neither is consistently better anyone hardcoding a single provider in 2026 is building a dependency the model is the least stable part of your stack rn

If your code can only work with Anthropic models, you will be stuck in a local maxima 🤷‍♀️ You need to rewire your code base for GPT 5.4 thinking It’s cheaper and more performant than Opus for some tasks

@johnrushx "skills will be the apps" sounds right but the distribution problem is identical actually app store had millions of apps and maybe only 200 that mattered the plugin store will have millions of connectors and 15 that anyone actually uses

The end game is that every serious tech company gonna build “general purpose agent” that can do anything (code, research, assist, etc). Same as the end game for mobile was a touchscreen with OS The skills/plugins/connectors/ will be the “apps” of the post-AI era for small teams

@fchollet true but the librarian metaphor undersells one thing: a librarian who can crossreference every paper ever written in 3 seconds is still pretty useful to the explorer the explorer just can't be the librarian too)

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.

@felixrieseberg @vite_js @tan_stack moving off SSR to static edge-served is such a clean win for an app like this

A small ship I love: We made Claude.ai and our desktop apps meaningful faster this week. We moved our architecture from SSR to a static @vite_js & @tan_stack router setup that we can serve straight from workers at the edge. Time to first byte is down 65% at p75, prompts show up 50% sooner, navigation is snappier. We're not done (not even close!) but we care and we'll keep chipping away. Aiming to make Claude a little better every day.

@AndrewYang already happening man i'm a freelance dev and my clients now casually ask "can't AI just do this?" before every project they're not wrong actually - it can do some of it but "feeling poorer" is exactly what happens when your skills get repriced overnight and nobody warned you

AI is going to make a lot of people feel poorer, not richer.

@callebtc commoditization is great for consumers and terrifying for anyone whose business model is "we have the best model" the moment 4 companies can all write decent code, the fight moves to tooling, ux, and who has the least annoying rate limits

LLM models are 100% turning into a commodity. Whether it's Codex, Claude, DeepSeek, Gemini – it matters less and less. Capabilities are a lot more similar now than a year ago. Most people won't care. New meta is price and speed. This is actually great news for consumers.

@Yuchenj_UW i use both and honestly the co-author thing is slightly annoying like not because i'm hiding ai usage but because my commit history now looks like a pair programming session with someone who never takes a day off and never argues back it's uncanny

I noticed something interesting: Claude Code auto-adds itself as a co-author on every git commit. Codex doesn’t. That’s why you see Claude everywhere on GitHub, but not Codex. I wonder why OpenAI is not doing that. Feels like an obvious branding strategy OpenAI is skipping.

@trq212 true, and the funny part is most teams still pretend it's a pure technical choice ]it's budget, latency, reliability, and whether future-you wants to debug a $40 workflow because someone was feeling visionary on a tuesday

an increasingly large part of the job of an engineer is deciding how much compute to spend on a problem

@sandeepnailwal "we built a mirror" is the right frame i've watched enough people project feelings onto llms after one good reply to realize most of this debate is just loneliness with better ux

LLM based AI is NOT conscious. I co-founded a company literally called Sentient, we're building reasoning systems for AGI, so believe me when I say this. I keep seeing smart people, people I genuinely respect, come out and say that AI has crossed into some kind of awareness. That it feels things, that we should worry about it going rogue. And i think this whole conversation tells us way more about ourselves than it does about AI. These models are wild, i won't pretend otherwise. But feeling human and actually having inner experience are completely different things and we're confusing the two because our brains literally can't help it. We evolved to see minds everywhere and now that wiring is misfiring on language models. I grew up in a philosophical tradition that has thought about consciousness longer than almost any other, and this is the part that really frustrates me about the current conversation. The entire framing of "does AI have consciousness?" assumes consciousness is something you build up to by adding more layers of complexity. In Vedantic philosophy it's the opposite. You don't build toward consciousness. Consciousness is already there, more fundamental than matter or energy. Everything else, including computation, is downstream of it. When someone tells me AI is "waking up" because it generated a paragraph that felt real, what they're telling me is how thin our understanding of consciousness has gotten. We've reduced a question humans have wrestled with for thousands of years to "did the output sound like it had feelings?" It's math that has gotten really good at predicting what a conscious being would say and do next. Calling that consciousness cheapens something that Vedantic, Buddhist, Greek and Sufi thinkers spent millennia actually sitting with. We didn't build something that thinks. We built a mirror and right now a lot of very smart people are mistaking the reflection for something looking back.

@ThePrimeagen this is how half the internet works now. random badge, confusing ui, zero explanation, and suddenly you're doing pr for a company you don't even work at

i don't work at x, i somehow have a badge. i use to be a "featured streamer" and i think this badge was recently updated to look like x, which makes me look like i work at x thank you for coming to my public service announcement, now get back to neovim you loser

@pvergadia the part nobody wants to say out loud is that ai makes bad devs feel faster and good devs feel weirdly guilty for not using it enough then prod goes on fire and suddenly everyone rediscovers the ancient art of understanding their own code

JUST DROPPED: Anthropic's research proves AI coding tools are secretly making developers worse. "AI use impairs conceptual understanding, code reading, and debugging without delivering significant efficiency gains." -- That's the paper's actual conclusion. 17% score drop learning new libraries with AI. Sub-40% scores when AI wrote everything. 0 measurable speed improvement. → Prompting replaces thinking, not just typing → Comprehension gaps compound — you ship code you can't debug → The productivity illusion hides until something breaks in prod Here's why this changes everything: Speed metrics look fine on a dashboard. Understanding gaps don't show up until a critical failur and when they do the whole team is lost. Forcing AI adoption for "10x output" is a slow-burning technical debt nobody is measuring. Full paper: arxiv.org/abs/2601.20245