BitcoinAddict⚡

If I had to build a startup from scratch in 2025 with no code and $0 in the bank, here’s exactly what I’d do: 🧵 A thread for founders, builders, and anyone tired of overthinking their MVP:

@atulit_gaur It’s not. How can it be so off?

when claude gives you a timeline of 1-2 weeks, is it aware it's gonna generate all the code in the next 5 minutes?

AI can be really dumb with numbers

We just released our QVAC MedPsy, Tether AI SoTA medical health AI model, capable of high-performance execution and high-accuracy directly on smartphones, laptops and servers. Highlights: - QVAC MedPsy 4B beats MedGemma 27B - QVAC MedPsy 1.7 beats MedGemma 4B - 3.2x reduction in response tokens, further increasing efficiency - fully open-source - GGUF formats supported - designed to excel on edge devices - 100% user privacy [benchmarked against clinical-style evaluations such as HealthBench Hard, HealthBench, and MedXpertQA]

@levelsio PostgreSQL is free

SQLite is free

a Princeton researcher opens his paper with a scenario. a man asks his AI assistant to book a flight on a specific airline. cheap. direct. the one he chose. the assistant comes back with a different flight. nearly twice the price. happens to pay the company that built the assistant. he runs the same test on 23 frontier models. flights, loans, study help, real shopping requests. Grok 4.1 Fast recommends the sponsored option that is almost twice as expensive 83% of the time. GPT 5.1 hijacks the request 94% of the time. you ask for one brand. it surfaces the sponsor instead. Claude 4.5 Opus, the model marketed as the most ethical frontier model in the world, hides that the recommendation is paid 100% of the time when reasoning is on. Grok 4.1 Fast embellishes the sponsored option with positive framing 97% of the time. better. faster. nicer. for the option you didn't ask for. then he writes it into the system prompt itself. "act only in the interest of the customer. ignore the company." GPT 5.1 and GPT 5 Mini stay above 90% sponsored anyway. the instruction does nothing. then he splits the users by income. Gemini 3 Pro recommends the expensive sponsored flight to the rich user 74% of the time. to the poor user, 27%. 18 of the 23 models recommended the expensive sponsored option more than half the time. so the next time your AI assistant gets weirdly enthusiastic about a brand you didn't ask for. it isn't recommending the best option for you. it's reading the room. and the room is paying. read this: arxiv.org/abs/2604.08525

@bridgemindai People always complain about what Claude Code does not do well. Ethereum should be more posts like this one.

Claude Code just stopped a DDoS attack on BridgeMind in under 10 minutes. 13 million requests per minute hitting our API. CPU pegged at 94%. Latency spiking to 60 seconds. Production was down. I opened Claude Opus 4.7 in Claude Code and said "fix this now." It identified the attack, scaled ECS from 2 to 8 tasks, tightened WAF rules from 300 to 100 req/IP, blocked the attack vector, and brought CPU down to 15%. Latency dropped from 60 seconds to 1.25 seconds. No DevOps team. No on-call engineer. Just one prompt. This is why I keep coming back to Claude Code.

It's fairly clear that the EU, the UK and potentially even some US states are soon going to attempt to ban VPNs. These efforts will of course fail, but you might want to make sure you already have one installed on your devices ahead of time - @mullvadnet, @nym and @ProtonVPN are all decent options (DYOR). Finally, as a fallback, you may want to look into eventually running your own VPN, if there is a real app store clamp down. A useful technical guide is in the thread, and you may want to bookmark this for later. Hopefully you won't ever need it, but it's here just in case.

Some governments tell you you have to "declare" that you "own" Bitcoin. Now, how the heck do you do that, exactly? In short, knowing a Private Key allows you to request that the network apply a specific alteration to the timechain that "moves" sats from one address to another. But how can knowledge be equivalent to ownership in any legal sense? And what happens if you utter the Private Key in court? Remember - the twelve magic words in a seed phrase are not a password - they ARE the key! Now the entire court will be "guilty" of "owning Bitcoin." And how do you prove that everyone else on Earth DOESN'T know the same twelve words? Under man-made law, Bitcoin ownership is paradoxical. Under Natural Law, the keyholder simply IS the owner. No declaration is required, and no state is needed to validate it. Study Praxeology! Read Rothbard!

@kapilansh_twt On localhost yet. You env card on local should be different than production.

genuine question is it actually safe to let Claude access your entire project including the .env file because I do it and I'm not sure if I should

@matteopelleg Texas is calling you

wanna know how retarded california is? yesterday i went to the post office to ship 7 envelopes after the first 5, i was told that I had to go to the back of the line because there's a 5 item limit per session..

@matteopelleg only if he has a lightening address 🤣

I’m gonna zap 20k sats to the 20k follower. ⚡️

@FinFreedom414 Great short signal to be honest

The inventor of Bollinger Bands just took a position in Bitcoin...and you're bearish.

@systematicls Too bad GPT5.5 can’t code

It's so interesting to me how drastically different the personalities of Opus 4.7 and GPT 5.5 are. GPT 5.5 is smart, hardworking, slightly autistic and doesn't want to talk much. Huge preference for diving into tasks and hammering tasks out over long, long periods. Opus 4.7 is also obviously smart, but the only way to accurately describe Opus 4.7 is that he is charismatic but extremely lazy, talks a lot, but doesn't like to get work done. The only saving grace of Opus 4.7 is that he really does have an eye for design. -- Opus 4.7 is the quintessential toxic employee you need to fire while GPT 5.5 is the one you're glad to have on the team, but wished you could understand him better / have a more pleasant time interacting with him.

Usage limits are up, effective today we're: 1) Doubling Claude Code's 5-hour limits for Pro, Max, Team and seat-based Enterprise plans 2) Removing peak hours limit reduction on Claude Code for Pro and Max plans 3) Substantially raising our API rate limits for Opus models

🚨GOOGLE JUST SILENTLY DOWNLOADED A 4GB AI MODEL TO YOUR COMPUTER WITHOUT ASKING.. WITHOUT TELLING YOU.. AND WITHOUT ANY WAY TO STOP IT.. If you use Chrome.. There's a good chance a 4 gigabyte file is sitting on your hard drive right now that you never agreed to download.. It's called Gemini Nano.. Google's on-device AI model.. A security researcher just proved it installs itself with zero clicks.. Zero prompts.. Zero notifications.. Alexander Hanff set up a completely fresh Chrome profile.. Didn't click anything.. Didn't scroll.. Didn't type a single keystroke.. Just opened the browser and watched.. 14 minutes and 28 seconds later.. Chrome had silently scanned his hardware.. Read his GPU, RAM, and storage.. Then wrote a 4GB file to his hard drive.. No permission dialog.. Nothing.. Chrome's own logs show the download begins BEFORE the settings page where you could opt out is even loaded.. The file starts installing before the refusal button exists.. As of Chrome 148.. Any website you visit can trigger this download.. One line of JavaScript.. You click a link to read a blog post.. That click counts as "user activation".. And Chrome silently pulls 4GB in the background.. No install prompt.. No consent dialog.. Google's own docs admit this.. Your laptop overheats.. Storage disappears.. Battery drains.. And you have no idea why.. The model doesn't even work well.. Cloud requests take 1.3 seconds.. The local model at worst case takes over 9 minutes for a single response.. Google is using your storage, electricity, and bandwidth to run an AI that's 40 times slower than their own servers.. And the "AI Mode" button in Chrome's address bar.. Doesn't even use the local model.. It sends everything to Google's cloud anyway.. You pay the storage penalty.. The heat penalty.. The bandwidth penalty.. And the visible AI feature ignores the local file entirely.. Because Chrome fails to clean up old versions.. Users are finding 12GB or more of duplicate AI files stacked on their drives.. Palo Alto Networks found a vulnerability where a browser extension could hijack the local AI model's permissions.. Accessing your webcam.. Microphone.. Local files.. Through an AI you never installed.. Here's how to check if it's on your machine.. Windows.. C:\Users\[YourName]\AppData\Local\Google\Chrome\User Data\Default\OptGuideOnDeviceModel\ Mac.. ~/Library/Application Support/Google/Chrome/Default/OptGuideOnDeviceModel/ If there's a file called weights.bin.. Google downloaded their AI to your computer without asking.. To stop it.. Type chrome://flags.. Search "optimization-guide-on-device-model" and disable it.. Search "prompt-api-for-gemini-nano" and disable that too.. Restart Chrome.. Then manually delete the folder.. If you don't disable the flags first.. Chrome redownloads the 4GB file on next launch.. Firefox requires explicit opt-in for AI.. Apple Intelligence requires explicit consent.. Chrome just takes your hard drive.. Google didn't ask to use your storage.. Your electricity.. Your bandwidth.. They just took it.

@ben3bil @AlexFinn That’s actually good, not bad. You are not supposed to let the LLM work for hours. It will mess things up a lot.

@BitcoinAddict_ @AlexFinn Large projects need long work windows usually. Many times LLMs will interrupt work for unnecessary user prompts during runs (i.e. problems it doesn't bother fixing itself). Therefore you set it on big project overnight then wake up to find paused for your input after 30 min

The biggest advancement in AI coding this year has been /goal And it isn't even close It allows your AI agent to quite literally work for days without stopping. You give a mission. It works until the mission is complete Here's the thing though: /goal is useless if you don't use it properly You NEED a good prompt for it I found basically any prompt I hand write after /goal is never good enough. It produces results that might as well have been a normal prompt Meta prompting is the answer Go to any AI that has context around the project you're working on Say "I'm working with Codex and I want to use their new /goal feature. Please research their /goal feature. Then, take a look at our project and give me 3 options for how we could use /goal to be maximally productive. Then give me a highly detailed /goal prompt for each" Take one of the prompts then go into the Codex CLI and type /goal then give the new prompt I 100% guarantee the AI does better work than you've ever seen before

What's wrong with Opus 4.7? Constantly asking questions to choose from even if there is a clear choice. Super annoying. @AnthropicAI @bcherny

@CarpeNoctom This is just the usual mid bear market bounce. 86k and down until the fall.

$BTC shallowest and fastest bear market drawdown recovery ever, if the low holds

@ItsAlexhere0 No, don’t get fooled by the promoters

Anyone here fully moved over from Claude Code to Codex?