Oishee Choudhury

[Anoy @Anoyroyc]

"what's the point of building in public if it just exposes your failures?"

the exposure is the point. it forces you to confront what isn't working and in doing so build better instincts.

for many who have shipped broken things and pivoted in front of everyone

[Anoy @Anoyroyc]

🚨ZAI just dropped an autonomous coding IDE with multi-agent collaboration for $18/month.. while GitHub Copilot still needs you to write half the code yourself..

the gap is closing way faster than anyone thought..

Western AI companies are about to get their asses kicked by teams that move like startups, not slow-motion corporations..

[Anoy @Anoyroyc]

1. Open your laptop. Ship something small. Do that for your momentum.
2. Talk to one user today. Do that for your product.
3. Write down what broke yesterday. Do that for your systems.
4. Close a tab. Focus. Do that for your output.
5. Repost. Do that for others.

[Anoy @Anoyroyc]

🚨Anthropic built a perfectly safe AI called Fable 5.. then the US government banned it for 3 weeks.. not because it was dangerous.. but because they couldn't verify if foreigners were using it..

the AI passed every safety test.. answered questions about recipes and homework.. couldn't make weapons or hack anything.. basically an expensive search engine with manners..

but export control laws don't care about safety.. they care about "dual-use technology".. since AI could theoretically be military.. every system needs export licenses..

here's the stupid part.. the government can't check users in real-time.. so they told Anthropic to shut down until compliance was sorted..

three weeks later.. after geo-blocking and verification dashboards.. they got permission to turn back on the AI helping kids with math homework..

meanwhile China released 12 new AI models.. Russia launched 3 military AI programs.. America spent a month ensuring our safest AI couldn't teach a foreign student about photosynthesis..

we're losing the AI ra wrong things.. protecting against imaginary risks while competitors build actual weapons..

the bureaucrats are fighting the last war while the next one builds itself

[Anoy @Anoyroyc]

Idk who needs to hear this but building something real requires showing up every day and shipping at least one thing, anything, that moves it FORWARD. stop asking how fast someone grew their product when you should be asking how consistent they were.

It's gonna take as long as it takes you to stop waiting for the perfect idea and start iterating on the one you have. Study your users.

Learn to build in public. Now go ship something.

[Anoy @Anoyroyc]

🚨 Meta built a helmet that reads your thoughts and turns them into text.. 61% accuracy.. improving fast.. they call it a medical breakthrough.. but that's not who really wants this..

they're pitching it for paralyzed patients.. but Meta isn't in medicine.. they're in surveillance.. they profit by knowing what you think before you type it..

same company caught listening through your phone.. tracking you everywhere.. building profiles on people who never signed up.. now they want inside your head.. literally..

here's what should terrify you.. 61% means 6 out of 10 thoughts already read correctly.. and Meta's AI improves weekly.. remember when their translation was garbage.. now it's near-perfect..

"medical breakthrough" is the headline.. the real buyers are governments wanting mind-reading interrogations.. employers screening workers.. advertisers targeting thoughts you haven't had yet..

they're not building this for hospitals.. they're building it because the next level of control isn't what you say or do.. it's what you think..

helmet today.. contact lens tomorrow.. your thoughts become their product forever

[Anoy @Anoyroyc]

Someone told me once "it's not that I don't want to ship." They're now stuck in permanent polish mode.

But they didn't start that way. Most builders don't become perfectionists out of nowhere, they become it bc they shipped too fast once and got burned.

[Anoy @Anoyroyc]

let me explain what Anthropic's "AI safety" really means.. they decide who builds what.. who gets access.. who gets reported to the government..

they want to be the bouncer at the door of the most important technology in human history.. using your fear to get the job..

here's how it works.. Anthropic creates "safety standards" requiring resources only the biggest companies have.. then they call themselves responsible while smaller builders get tagged as "reckless"..

same playbook tobacco used.. fund the research.. control the story.. become the expert on your own product's dangers..

meanwhile actual safety gets worse.. fewer hands don't mean safer AI.. they mean AI that serves fewer people..

they're not AI.. they're protecting their spot at the top.. wrapping it in language that makes you thank them..

the most dangerous AI isn't unregulated.. it's the one that regulates itself while calling everyone else unsafe..

[Anoy @Anoyroyc]

The longer you build the more you need to:

- Ship daily
- Talk to users
- Ignore feature requests (mostly)
- Learn to say no
- Automate the boring stuff
- Rest before you're forced to
- Build with people smarter than you

Anything else you'd add?

[Anoy @Anoyroyc]

🚨 China just built an AI that matches Claude Mythos.. Anthropic's most restricted system.. so dangerous they won't let anyone touch it..

Claude Mythos sits in a vault.. tested on classified penetration tasks.. locked away because what it can do is too risky for anyone outside to see..

China matched it anyway.. same zero-day exploits.. same defensive systems cracked.. they reverse-engineered America's most guarded AI without ever seeing the code..

if China can match our most locked-down cyber AI.. they can blow past everything we're actually running.. every firewall.. every security tool was built for human-level threats..

both sides now have superhuman cyber weapons.. and the systems protecting power grids.. hospitals.. financial networks.. were built for a different world..

China didn't just build an AI.. they proved the security arms race is already over.. and nobody told the people keeping the lights on..

[Anoy @Anoyroyc]

The most important shift I've made as a builder is stopping the wait for the "right time."

Ship. Test.

Break. Rebuild.

Move from planning to doing.

The product won't build itself.

[Anoy @Anoyroyc]

🚨 Anthropic's Claude Mythos 5 just got banned globally for national security.. but 100 "vetted American organizations" still get access..

the AI that's supposedly too dangerous for public use.. too risky for regular developers.. but somehow safe enough for corporations that already control everything..

while you're stuck with rate limits.. Goldman Sachs gets unlimited access.. while indie developers get error messages.. Raytheon gets priority support..

they didn't ban it because it's dangerous.. they banned it because dangerous tools aren't for people like us..

the same pattern every time.. create something world-changing.. call it too risky.. then hand it to the people who already have all the power..

thy.. it's about making sure the next generation of AI serves the last generation of power..

they're not protecting you from the technology.. they're protecting the technology from you..

[Anoy @Anoyroyc]

If someone's only response to a failed product is "you just need more marketing," you are not speaking to someone who understands what building actually requires.

[Anoy @Anoyroyc]

Alibaba just proved you don't need to spend billions building AI.. you just steal it..

they created 25,000 fake accounts and had 28.8 million conversations with Claude..

it's called 'distillation attack'.. you train your model with responses from your competitor's model..

Anthropic spent years and hundreds of millions building Claude..

Alibaba copied it in 45 days for basically free.. and it wasn't just Alibaba..

DeepSeek, Moonshot AI, MiniMax all did it.. why build when you can steal..

[Anoy @Anoyroyc]

3 Truths About Building:

1. Building isn't about shipping fast.

It's about solving real problems.

2. The best products succeed because they make users feel understood.

3. You don't have to be technical to build, but you have to keep building to get good.

[Anoy @Anoyroyc]

🚨 OpenAI just told Nvidia to go fuck itself.. built their own AI chip in 9 months.. called it Jalapeño.. designed to replace Nvidia's GPUs for ChatGPT..

OpenAI was spending billions on Nvidia chips.. every ChatGPT query running on hardware Nvidia controlled the price of.. and they kept raising prices because they could..

so OpenAI said fine.. we'll build our own.. 9 months.. not 9 years..

here's what should scare every tech monopoly.. they built it to kill Nvidia's advantage.. optimized for transformer models.. large language models.. the exact workloads that made Nvidia billions..

now OpenAI controls their own destiny.. no Nvidia markup.. no chip shortages.. no hostage negotiations..

this is what happens when you charge toosmart people can build themselves.. every AI company is watching.. Nvidia's monopoly just got shakier..

the chip is called Jalapeño because things are about to get spicy for Nvidia's stock..

[Anoy @Anoyroyc]

it is relatively easy to ship a feature you've seen someone else build.

it is extremely hard to build something new, with no playbook, when you have no idea if it will land.

founders who do it anyway deserve way more credit than they get.

[Anoy @Anoyroyc]

the government is begging Meta to hand over AI models for security reviews..

OpenAI, Google, Microsoft signed up.. 30 days to find flaws before release.. sounds responsible..

except it's voluntary.. companies pick what gets reviewed.. they're grading their own homework..

we're getting security theater.. companies hand over their safest models.. get a government stamp.. everyone pretends the problem's solved..

here's the kicker.. Mythos found critical flaws in the government's own AI.. the agencies demanding reviews can't secure their own models..

real threats build in silence.. while we clap for the ones who showed up..

we're checking IDs of people who want to be checked.. while actual risks walk through the back door..

[Anoy @Anoyroyc]

Your first idea is for exploring.
Your first build is for breaking.
Your first launch is for learning.
Your first traction is for keeping.

The tragedy is most builders never make it past the breaking phase.

[Anoy @Anoyroyc]

everyone's arguing about which AI model is best while Elon just became the landlord for all of them..

$28 billion a year from three deals because every AI company realized they need his rockets to run their code..

you can build the smartest model in the world but it doesn't matter if you can't plug it in anywhere..

while everyone fought over features.. he built the foundation