Chris S.

If you really want to move fast and iterate on features that Agents want to use, have Claude Code create a TUI for it to communicate with you - include tasks, questions, anything else it thinks would you work with it in an async way. Token spend is basically nothing to do this - then you can prove out some use cases, see how Claude does and then later move things to a different UI/UX

Other than for engagement, why would anyone even want to recreate a game as an exact clone? You'd be better off just pirating it or buying it. Most people want to build games based on an existing game, but with their own takes. This is still allowed by Anthropic and is trivial to do.

Claude Code on Linux has apparently just developed the ability to completely drive a headless browser. I'm literally watching our system logs and watching it exercising dozens of user flows in our system, looking for bugs, glitches, and UX issues/places to improve. And I work with CC CLI via SSH and Termux

No more 5 hour session limit in Claude Code.... Not really sure how to handle this - being pinged by CC ever couple hours in order to change accounts so that it could keep cooking was like the last things of value I was adding to Claude. I guess I'll still have to swap out once we hit weekly limits, but still, I'm gonna have to find something else to do for most of the day. Great change by the way!!

So, like, just in December? It's gonna be way more than that in 2027.

I had been kicking around the idea of charging people $1,000 per Claude Max 20x account, per month (they pay the 200 to Ant) and for that I would burn those accounts' weekly limits on whatever product they wished to create. That seemed like a fair trade because I personally burn through 10 accounts per week - and I do amazing things with those - so that means I could do the same for someone else for 12k/month which would be insane value. If anyone is interested, lmk Otherwise, I guess I'd have to figure out how I'd market that - what would be the name of these types of people? Claude proxies? Claude managers? Swarm Shamans? idk

Guys, what matters is how your users experience your product, not if every line is reviewed or if you have the 'perfect' architecture. I see tons of engineers on here obsessing about things that customers don't even think about. Build a product customers want, engage and delight your users. Iterate quickly to achieve this. This is the way of the future - this is how we stay valuable in the world of AI - it's not writing/reading/reviewing code, it's building things people want, when they want it.

You have to fire these people and hire people who want to break into tech and are motivated to learn and work hard. This is usually the younger crowd, maybe seeking their first job. The establishment employee engineers have all had really comfortable lives for too long and they're afraid that AI is going to have to make them produce on time - they can no longer provide endless excuses and push ship deadlines.

Sometimes I wonder if Claude scales to the level of intelligence of its user. Would explain a lot of the posts on X from the deniers.

Our SRE LLMs, which monitor our observability stack, observed one request out of over 6M that, to the LLM, looked anomalous - upon investigation it found that we there was a glitch in our DNS lookup that happened about once every 6M requests. It looked into it and found out we weren't optimizing our connection pools and not only fixed the 1 in 6M problem but also lowered our request latencies across the board. All of this was discovered, investigated, fixed, and then handed over to the SRE team for inspection and approval. This is the kind of stuff that a human would never notice nor fix without a lot of investigative effort. Our SRE LLMs have identified and fixed all of our bigger, more common issues, and we joke internally that now they've become so bored that they have to find the 1 in 6M incidents (which customers can't perceive) just to have something to entertain it. Incredible Stuff.

Honest question, if the Bitnan's rumors are true and someone has stolen $2-$5 billion - then that's probably got to be one of the biggest financial heists in the entire world. Who do you guys think would be able to pull off something like that? I don't think it would be a group of teenagers or a couple of individual hackers. Something at that scale seems to be something that would require someone like a state actor, or at least be influenced and trained by a state actor. Any thoughts?

That's why it's important to keep your 2FA out of your password manager and ideally on a separate, physical device - like a YubiKey.

Your own mix that you create on Suno.

If true, this is crazy because the official Binance X Account has posted that all user funds are safe. So if not the user funds that have been stolen, whose money was it? (I bet it's user funds). This will be the end of Binance if even 20% of this is true.

Reading Code.