Rohit
@rohit4verse • 23,381 subscribers
Engineer who builds, solves, and ships | FullStack + Applied AI | Agentic AI |
Shorts
Videos
Anthropic CEO: "we got seven more months." the bet was a $1B one-person company by end of 2026. two-person AI companies already crossed $1B, one-person companies are past several hundred million. not everyone can make a $1B company. but a $10K MRR AI agent company is on the table. this guy dropped the exact roadmap. Bookmark this and start this weekend.
Rohit1,028,441 görüntüleme • 26 gün önce
SaaS is dead. Anthropic's CEO said it on camera. So who survives? The people who own the workflow layer that ate it. I wrote the build path for that pivot in January. Bookmarks went from 5k to 27K while everyone argued whether he meant it literally. 5 projects. Beginner to master. Read it before the wrapper crowd catches up.
Rohit318,218 görüntüleme • 28 gün önce
this is how the founder of obsidian actually takes notes in his own app. most users get this wrong. barely any folders. heavy internal linking. categories live as properties on the note itself. article below teaches how to build personal knowledge base. video above is the underrated masterclass.
Rohit231,114 görüntüleme • 21 gün önce
Sam Altman said on camera at Stripe Sessions that we are in an insane area, stuck with old devices and old operating systems. The OS and the browser are built for humans clicking. Both get rewritten around agents that do not click. Two years deep building agents. Sharpest framing I have heard a frontier lab CEO put on the next interface. ↓ Read it before the gap locks in The part nobody is sitting with: >the chat window is the old interface >the new one is an agent in a loop with your machine >the gap between people prompting and people running agents is forming right now >in six months it will be permanent Open Claude, type a question, copy the answer, paste it into VS Code. The average dev runs this loop every day. They think they are using AI. That is 5% of what these tools can do. The people pulling ahead do not prompt. They build harnesses. They run agents in loops. Same model. Different output. Full 2026 agent playbook in the article below here:
Rohit120,392 görüntüleme • 13 gün önce
Roy lee(Cluely CEO) at NYU offered $500 cash to anyone in the room who had ever shipped a project on linkedin. almost nobody qualified. most engineers in 2026 are still in that room. Don't be that engineer. Pick a project from the 2026 ai engineer roadmap, build this weekend. publish by next.
Rohit362,747 görüntüleme • 1 ay önce
Hermes Agent just passed OpenClaw as the most-used open agent on OpenRouter. It writes its own skills from experience and remembers every session. New to it? Watch the 5-min explainer, then read this guide to run your own. Best hour you’ll spend on agents this month.
Rohit33,385 görüntüleme • 4 gün önce
>be Naval Ravikant >spawn in New Delhi >poor immigrant family >father leaves shortly after they land in Queens, New York >mother works menial jobs by day, night school after >age 9, alone in a country you don't understand >no friends >no safety net >no connections >the library becomes your entire world >get into Stuyvesant High School >yes, that Stuyvesant >the public school that produced four Nobel laureates >one entrance exam changes the trajectory of your life >go from blue collar to white collar in a single move >graduate and land Dartmouth >double major >computer science and economics >pay your way through by washing dishes, delivering newspapers, tutoring, fixing computers >move to Silicon Valley with nothing but a degree and a bet on yourself >join Geneia@home Network >watch a $20 billion company go to zero in the dot-com crash >work on Intrinsic Graphics >the thing that eventually becomes Google Earth >co-found Epinions in 1999 >a consumer review site before Yelp and TripAdvisor existed >raise $45 million in venture capital >get screwed by your own co-founder and VCs >they hide the company's real value during a merger >the company IPOs at $750 million >you walk away with $0 >not a typo >zero >sue Benchmark Capital and August Capital >everyone in Silicon Valley calls you radioactive >one VC tells the press you'll never work in the valley again >settle the case >learn the game from the inside out >instead of quitting, you decide to rewrite the rules >start Venture Hacks in 2007 >a blog that tears the veil off VC term sheets >give founders the playbook that VCs never wanted them to have >launch a $20 million fund called Hit Forge >back Twitter before anyone cares >back Uber before anyone believes >back Stack Overflow, Notion, Postmates, Opendoor, Yammer >turn Venture Hacks into AngelList in 2010 >50 angel investors >$80 million committed in year one >100 new startups signing up per day >build the for founders and investors >do what LinkedIn tried and failed to do >actually get people to transact >realize US securities law is blocking everything >fly to Washington DC >spend six months lobbying Congress >rally 5,000 investors and entrepreneurs for an online petition >call in 100 favors >get the JOBS Act signed into law by Barack Obama in 2012 >single-handedly open startup investing to ordinary Americans >the entire equity crowdfunding industry exists because of this >AngelList hits $4 billion valuation >over $3.5 billion invested through the platform >200+ unicorns funded >spin off Product Hunt, Republic, CoinList >companies like Neuralink and Rippling get backed through your Spearhead fund >co-found MetaStable Capital in 2014 >a crypto hedge fund backed by Andreessen Horowitz and Sequoia >bet on Bitcoin and Ethereum before it's fashionable >earn a penny on every Uber ride taken on the planet >invest in 200+ companies across your career >drop a 40-tweet storm on May 31, 2018 >"How to Get Rich (without getting lucky)" >it goes more viral than most product launches >turns into a podcast series, then a global movement >Eric Jorgenson compiles your tweets and interviews into The Almanack of Naval Ravikant >it becomes a perennial bestseller >you didn't even write the book >your ideas were so good someone else did it for you >go on Joe Rogan, Tim Ferriss, Shane Parrish, Chris Williamson >every episode becomes a top-10 all-time listen >millions of people re-read your tweets like scripture >you become the most quoted man on the internet who isn't dead >launch Airchat in 2023 >voice-first social media with AI transcription >because you think text-only platforms made us forget humans can get along >here is what Naval actually taught the world >you're not going to get rich renting out your time >own equity or stay a renter forever >specific knowledge is the stuff that feels like play to you but looks like work to others >leverage is a force multiplier for your judgment >code and media are permissionless leverage >you don't need anyone's approval >10,000 iterations is not 10,000 repetitions >one is mastery, the other is a treadmill >if you can't decide, the answer is no >when two choices look equal, pick the harder one short term >stress is an inability to decide what's important >desire is a contract you make with yourself to be unhappy until you get what you want >inspiration is perishable >act on it immediately >the three big decisions: what you do, where you live, who you're with >people spend years optimizing careers but pick partners and cities on autopilot >not optimizing for wealth >optimizing for sovereignty >win the game fast enough that you get to stop playing
Rohit485,379 görüntüleme • 2 ay önce
I thought I was a Claude Code power user. Then the creator of Claude Code(Boris Cherny) dropped this thread and I realized I've been using maybe 30% of the tool. >Scheduled automation loops. >Hooks into the agent lifecycle. >Git worktree parallelism. >Fan-out across hundreds of agents at once. >Coding by voice. This thread is mandatory reading if you're building with it.
Rohit412,458 görüntüleme • 2 ay önce
a16z just dropped the billion-dollar opportunities in AI for 2026. three partners. three theses. same underlying bet. Marc Andrusko: the prompt box is dying. next-gen apps observe what you're doing and act on your behalf. TAM shifted from $ 400B software spend to $ 13T labor spend. market got 30x bigger. Stephanie Zhang: stop designing for humans. start designing for agents. agents read every word on the page. visual hierarchy stops mattering. GEO is the new SEO. Olivia Moore: voice agents ate the phone in 2025. healthcare, banking, recruiting, 911 calls. voice AI beats humans on compliance every single time. some companies now slow their agents down to sound human. every thesis converges on the same layer. the harness around the model is where the leverage compounds. full breakdown of how the shift happened below.
Rohit193,782 görüntüleme • 1 ay önce
LangChain's co-founder: "memory can be a real moat." the architecture that builds this? four markdown files and a cron job. someone open sourced the full stack. memory layers, self-rewriting skills, a dream cycle that runs at 3am. yes, you should read every word of this.
Rohit219,580 görüntüleme • 1 ay önce
![andrej karpathy spent two hours teaching one thing: tokens are the atom of llms. tokenization is at the heart of every llm weirdness you've ever debugged. [watch the 15-min clip below. then run the 7-day playbook] ↓ save this before everyone copies it learn how the tokenizer works. understand how your llm actually consumes input. then run the engineering roadmap that took one production agent from $4,800/mo to $620/mo in 7 days. 87% reduction. no model swap. no framework migration. no quality drop on the eval set. token cost in 2026 is an engineering discipline. every line of your system prompt is rent you pay forever. what was eating the budget: → a single forgotten cron job ate 47% of one team's bill. they turned it off on a tuesday and the bill dropped before they wrote any optimization code. → anthropic ships a 90% discount on cache reads. one config line, cache_control ephemeral, break-even after one hit. most teams cache the volatile parts of the prompt and watch their hit rate sit at 12%. → one production agent went from 14,500 tokens of context overhead per turn to 850. a 94% drop. output quality held within 2% of the uncompressed baseline. → 60% of agent calls are haiku-tier work running on opus rates. classify the task first. pick the model second. → retry loops are the silent killer. no MAX_STEPS bound, one bad search query, $14 burned in a single session. one team traced 38% of their bill to this single pattern. karpathy gave you the atom. the playbook below gives you the harness. watch the lecture. read the playbook ↓](https://image.24vids.com/tw-2056429557901877579/amplify_video_thumb/2056429198072553472/img/PaSBAZOMOB1hWDUi.jpg)
andrej karpathy spent two hours teaching one thing: tokens are the atom of llms. tokenization is at the heart of every llm weirdness you've ever debugged. [watch the 15-min clip below. then run the 7-day playbook] ↓ save this before everyone copies it learn how the tokenizer works. understand how your llm actually consumes input. then run the engineering roadmap that took one production agent from $4,800/mo to $620/mo in 7 days. 87% reduction. no model swap. no framework migration. no quality drop on the eval set. token cost in 2026 is an engineering discipline. every line of your system prompt is rent you pay forever. what was eating the budget: → a single forgotten cron job ate 47% of one team's bill. they turned it off on a tuesday and the bill dropped before they wrote any optimization code. → anthropic ships a 90% discount on cache reads. one config line, cache_control ephemeral, break-even after one hit. most teams cache the volatile parts of the prompt and watch their hit rate sit at 12%. → one production agent went from 14,500 tokens of context overhead per turn to 850. a 94% drop. output quality held within 2% of the uncompressed baseline. → 60% of agent calls are haiku-tier work running on opus rates. classify the task first. pick the model second. → retry loops are the silent killer. no MAX_STEPS bound, one bad search query, $14 burned in a single session. one team traced 38% of their bill to this single pattern. karpathy gave you the atom. the playbook below gives you the harness. watch the lecture. read the playbook ↓
Rohit72,715 görüntüleme • 17 gün önce
Harrison Chase(LangChain CEO) just walked through four ways to give an agent memory. All four assume the model is still holding the right tokens. It isn't. At token 4,096 the cache ran a silent eviction nobody wrote. The user's name was in that batch. First founder to write the eviction policy ships a 100B agent that remembers a person.
Rohit108,195 görüntüleme • 1 ay önce
Anthropic published an article how to build a long-running agent. > Four compaction strategies, cheapest to most expensive. > Disk-backed task list locking state across parallel agents. > CLAUDE.md hierarchy persisting memory across sessions. Full implementation here:
Rohit136,746 görüntüleme • 1 ay önce