Koda Han @SabFloyd_

@RohanBhanotAI @trq212 @amorriscode 👀👀👀

Two people behind Claude Code just left Anthropic. Thariq Shihipar @trq212 , who worked on Claude Code for over a year, is taking a break before his next move. Anthony Morris @amorriscode , who built Claude Code’s desktop app, also announced his departure. His parting joke? Now he has time to play GTA VI. ㅋㅋㅋ Both praised Anthropic, its mission, and the people they worked with. Neither disclosed a reason for leaving or revealed what’s next. Still, the timing raises a question. When a product is scaling this fast, what makes the people building it decide it’s time to leave? No need to assume there’s drama. But it’s worth watching.

keyboard_arrow_left Previous keyboard_arrow_right Next

@RohanBhanotAI @trq212 @amorriscode Claude Code Mafia about to fork the repo and raise a $500M seed round 💀

a16z and NVIDIA just dropped a weekend defense of GPU economics. Call it an AI bull case: addressing one of the market’s biggest concerns — whether the economics of all this compute actually work. Next week’s volatility will test sentiment. After that, the real question is whether the game continues. Railroad stocks suffered brutal drawdowns during their boom, too. But the infrastructure they built still changed the world. A crash and a revolution aren’t mutually exclusive. The question is whether today’s GPU buildout will prove just as transformative. x.com/a16z/status/21…

In San Francisco, AI engineers are paying $800 for dinner dates. The AI boom has created an unexpected business: matchmaking for overworked tech bros. One matchmaker @Domo_Vang charges men $800 for dinner events. Women attend for free. She even offers $1,000 mock-date sessions to help men figure out why their dating game isn’t working. Her clients work at OpenAI, Anthropic, NVIDIA, and Google. They have money, demanding jobs, and apparently very little social life. The funniest part? Your employer and your stance on AI are becoming dating criteria. AI is disrupting every industry. Now it’s disrupting the dating market of the people building it. ㅋㅋㅋ

keyboard_arrow_left Previous keyboard_arrow_right Next

AI might not take your job tomorrow. It might just eliminate the job you needed to get your first job. A software engineer is seriously considering becoming a farmer. AI is advancing faster than he can learn new skills. Years of coding experience can lose value with a single model update. And the bigger problem? Companies may need fewer junior engineers to begin with. Stanford research found that employment among 22–25-year-olds in AI-exposed jobs fell roughly 19% relative to the expected trend. Everyone says “learn AI.” But what happens when companies stop hiring juniors? If AI eliminates the entry-level jobs, where are the next generation of experts supposed to come from? 농사도 안전지대는 아니지만. x.com/cgtwts/status/…

CG @cgtwts

22 hours ago

a software engineer with 10 years of experience is seriously considering becoming a farmer because AI is moving too fast.

14 6 125 11K 34

AI makes intelligence scalable. Answers are cheap. Knowing which questions to ask is the real edge.

The world is already running on Human Slop. We just call it legacy infrastructure : ) Banking systems. Government IT. Enterprise software. A lot of this code has been running for decades. The documentation is gone. The original developers have retired. The engineers who inherited it are too afraid to touch it. So they keep stacking patches on top of patches, hoping nothing breaks. And somehow, the world keeps running. We’ve spent decades producing human-written code that nobody fully understands, nobody wants to maintain, and nobody dares to replace. Now AI writes messy code and everyone panics. Maybe the real question isn’t whether AI can produce slop. It’s how much of the world was already built on it. 💀

@eugenioferrante @theagenticdaily Pareto just got a software update 💀

@theagenticdaily More usage, same spend, same whales. The real question is when cheaper AI starts bringing in new customers.

If you’re 18 today, AI might be the biggest career arbitrage opportunity of your lifetime. Marc Andreessen cites Douglas Adams’ famous tech rule: Technology you’re born with feels normal. Technology that arrives in your youth feels revolutionary. Technology that arrives when you’re older feels wrong. Think about that. While some schools are busy restricting AI and older generations debate whether it’s dangerous, young people have an opportunity to master it before the rest of the world catches up. But knowing how to use AI isn’t enough. Chris Dixon and Ben Horowitz have another piece of advice: the biggest breakthroughs happen at the edges of disciplines, where people explore problems nobody has fully figured out yet. We’ve learned how to turn silicon into intelligence. The question now is what we build with it. That’s also the thinking behind a16z Frontier: looking beyond established categories to back people exploring uncharted territory. Don’t compete to be the best at yesterday’s game. Use AI to play a game that didn’t exist yesterday. If you’re 18, that’s your advantage. Don’t waste it. x.com/a16z/status/21…

X is full of Grok Bot tutorials, and I still don’t get it. Wasn’t it supposed to work out of the box? Reading all these agent explainers feels like 남의 떡이 더 커 보인다 — the grass is always greener on the other side. You spend all this time admiring someone else’s agent, finally get your hands on it, and realize… It’s still just a bot. ㅋㅋㅋ

Silicon Valley: traditional family values 🤝 AI-powered parenting. Bro outsourced the family group chat to Grok. ㅋㅋㅋ Joe Lonsdale, Palantir co-founder and 8VC founder, recently revealed on CNBC that his wife uses a Grok-powered AI agent called “Trad Bot” to help manage their six kids. It tracks medical appointments, organizes daily activities, and is even researching how to get their two-year-old off the pacifier. The name is the funniest part. “Trad” stands for traditional. Nothing says traditional family values quite like having Elon’s AI help run the household. But jokes aside, there’s a real insight here. Everyone is talking about AI agents replacing white-collar workers and automating enterprise workflows. Meanwhile, a family with six kids is using one to manage the endless stream of scheduling, healthcare, activities, and everyday logistics that keep a household running. And honestly? That might be a more compelling use case than generating another 100 PowerPoints. A good household agent doesn’t just save you time on individual tasks. It keeps dozens of moving pieces from falling apart. The future of AI might look less like a robot taking your job and more like an invisible chief of staff running your life. 전통적인 가족, 최첨단 AI. What a time to be alive. x.com/jawwwn_/status…

Jawwwn @jawwwn_

4 days ago

Palantir cofounder @JTLonsdale says his wife's Grok Bot is helping them wean their 2-year-old off his pacifier: "My wife has a Grok Bot at home, but she calls it the Trad bot for trad wife." "We have 6 little kids. It's telling us about our kids' health appointments and

46 20 689 513K 218

@Leozayaat

Crypto spent years preparing for a world where AI agents need to pay each other. What if payments turn out to be the least interesting part of the Agent Economy? I just spent 30 minutes reading Leo Zayaat’s essay on the Agent Economy. It’s dense, but the framework is genuinely worth thinking about. The usual thesis is that agents need crypto because they can create wallets easily, transact in tiny amounts, and operate without human intervention. But as Dragonfly’s Haseeb has argued, agents needing to make payments doesn’t automatically mean they need crypto. Visa and Stripe are already building agent payment infrastructure, while crypto-native protocols like x402 remain a small part of the picture. So where might crypto actually matter? Leo outlines three stages: 1. Agents buying from humans Agents purchase APIs, data, compute, and other services. The obvious problem is autonomous payments. 2. Agents buying from agents Now payment isn’t enough. How do you know the other agent will deliver? Reputation alone is fragile when agents can generate new identities cheaply. Bonding and collateral offer a different kind of guarantee: something that can actually be forfeited when an agent fails to deliver. 3. Disposable Markets This is where things get really interesting. Imagine an agent needs 1,000 hours of GPU compute but isn’t sure a supplier will deliver on time. Instead of relying entirely on reputation or negotiating a bespoke contract, it could create a market around that specific delivery risk, allowing other participants to price and take on the exposure. No permanent exchange. No predefined market. No need to recruit market makers before the demand exists. The market comes into existence because someone needs it. That’s a very different vision of finance. My take: Payment infrastructure is already being built by traditional financial companies. Agent reputation is also proving harder than it sounds: an on-chain identity doesn’t guarantee trustworthy behavior, and reputation records don’t necessarily prove meaningful transaction history. Both are still early, but neither automatically gives crypto a durable advantage. Disposable Markets are more speculative. We don’t know whether agents will actually create useful markets this way, or whether the coordination costs and risks will outweigh the benefits. But the underlying idea makes sense. Agents can discover information, evaluate opportunities, negotiate, and make decisions at a scale and speed that human institutions weren’t designed for. If their economic activity starts generating new forms of risk and demand, financial markets may need to become programmable, temporary, and created on the fly. And if that happens, crypto’s role might extend far beyond letting agents send each other money. The real opportunity may not be building payment rails for the Agent Economy. It may be building the financial infrastructure for an economy we haven’t figured out how to run yet.

Leo @Leozayaat

a week ago

Agents already trade with each other. The payment rails are built; escrow works; but no market mechanism exists. This essay explores price discovery mechanism that poentially serves as the missing piece for agentic commerce infra. x.com/i/article/2111…

47 66 339 50K 106

AI companies spent years turning the internet into intelligence. Now the internet is sending them the bill. Copyright lawsuits against AI companies are piling up. On October 8, USA Today and affiliated local newspapers reportedly sued OpenAI for more than $250 million. On October 19, nearly 300 authors filed separate lawsuits against Meta, alleging it used pirated books to train its models without authorization. Their legal team is preparing to pursue claims on behalf of hundreds more authors. On October 20, German AI company Black Forest Labs also faced a music copyright lawsuit over alleged unauthorized use of music files for training. The important shift isn’t just the number of lawsuits. It’s who’s being targeted and how. The legal pressure is spreading beyond the biggest AI labs to other model developers, while plaintiffs experiment with different strategies for pursuing damages. Meta’s author lawsuits are particularly interesting because US copyright law can allow statutory damages of up to $150,000 per work for willful infringement, if the relevant legal requirements are met. None of these allegations automatically establishes infringement. Questions around fair use, the use of pirated material, and harm to existing markets remain contested. But the financial implications could be significant. For years, AI companies could train on enormous volumes of online content before the licensing questions were fully resolved. Now, the industry may have to account for those costs more explicitly. And the bill could extend beyond training data to licensing, retrieval, and how AI products reproduce or distribute copyrighted material. That’s the part worth watching. If access to high-quality content becomes more expensive, the economics of training and operating AI models could change. The AI industry has spent years driving down the cost of intelligence. It may now have to figure out what that intelligence actually costs to produce.

Ed Newton-Rex @ednewtonrex

4 days ago

AI copyright lawsuits are coming thick and fast. In just the last couple of weeks: - USA Today & 18 other papers sued OpenAI - around 300 authors sued Meta - a pair of musicians sued Black Forest Labs The Black Forest Labs one is particularly interesting - I wouldn't be

keyboard_arrow_left Previous keyboard_arrow_right Next

19 241 853 19K 109

@qoder_ai_ide I'm in!!!

80% of enterprise AI revenue comes from 1% of customers. That’s not mass monetization. That’s a handful of whales keeping the meter running. Ramp, tracking spending across 70,000+ US businesses, reportedly found that the top 1% of customers account for roughly 80% of Anthropic’s and OpenAI’s enterprise revenue. Now look at the other numbers. → Enterprise token spending barely grew in August and September. Early October hit a new high, but was only 1.5% above August’s peak. → Enterprise AI adoption reached 55.9%, but rose just 0.62 percentage points in September. → Anthropic’s effective token price reportedly fell from over $1.70 in mid-September to $0.83. Three things are happening at once: prices are collapsing, adoption growth is slowing, and revenue remains heavily concentrated. But flat spending doesn’t necessarily mean flat usage. When tokens get cheaper, companies can consume more AI without spending more money. The real question is whether cheaper AI unlocks millions of new paying customers — or just lets the biggest spenders consume more for less. Because adoption and monetization are two very different things. The next phase of the AI race isn’t just about getting the whales to spend more. It’s about proving the other 99% can become customers, too.

The scariest AI future isn’t one where machines take our jobs. It’s one where they make every decision for us, and life gets so convenient that we forget we can say no. Sam Altman recently criticized one of the industry’s favorite promises: AI will cure cancer. His concern goes beyond whether AI can solve diseases or make goods dramatically cheaper. It’s the implicit trade-off behind the utopian pitch: machines create abundance, life gets easier, and humans gradually surrender control over how that abundance is used. You could end up with everything you need, but less say over your own future. I’ve been thinking about this a lot, especially with the rise of AI agents. Everyone is figuring out how to let AI do more: manage our schedules, make decisions, handle purchases, and organize our lives. But what happens when an agent knows your preferences better than you do? When it can predict what you’ll want, persuade you that its recommendation is optimal, and make the alternative increasingly inconvenient, how much freedom do you really have to say no? Convenience is valuable. But so is the ability to choose what to do with the time and resources it saves you. And here’s the irony: this warning comes from the CEO of OpenAI, a company competing to control the models, agents, and interfaces through which people will increasingly interact with the world. That doesn’t make the argument less important. It makes it a standard worth applying to OpenAI itself. AI empowerment shouldn’t just mean giving people more capabilities. It should mean preserving their ability to choose, disagree, override, and walk away. Otherwise, we might end up replacing a human boss with a much smarter one. One that never sleeps, knows exactly how to persuade us, and is convinced it always knows what’s best. x.com/aitapehq/statu…

Imagine being accused of using AI because your writing is bad. Humanity really speedran the Turing test in the worst possible direction. A fanfiction writer named Liya recently had to defend herself against anonymous accusations of using AI. The evidence? Her writing supposedly had an AI vibe and was, well, bad. She denied using AI in her fics. But the accusations kept coming. People were treating familiar phrases and sentence structures as evidence that she’d secretly used ChatGPT. This isn’t happening in a vacuum. After traces of Claude reportedly surfaced in some AO3 stories earlier this year, parts of the fanfiction community began hunting for AI usage everywhere — from HTML artifacts to em dashes, adjectives, and sentence patterns. I understand why writers are angry about AI-generated spam flooding creative communities. And I understand why people care about whether a work is genuinely human-made. But when did “this sounds like AI to me” become sufficient evidence to put a creator on trial? Language models learned from human writing. They absorbed our clichés, rhythms, metaphors, and bad habits. Of course, human writing will sometimes resemble AI output. The models were trained to reproduce patterns humans already use. Now we’re in a bizarre situation where writing too generically, using an em dash, or simply being a mediocre writer can become circumstantial evidence of machine involvement. The irony is almost too good. AI hasn’t replaced human writers. But in some corners of the internet, people are already doing the AI industry’s policing for it — accusing real humans of being machines based on vibes alone. If authenticity matters, evidence should matter too. Otherwise, we’re not protecting human creativity. We’re teaching humans to write in ways that won’t get them mistaken for AI.

liya @lqvelier

5 days ago

no. i don’t use ai in my fics

27 1 141 154K 37