Mark M.

We won $40,000 at hackathons without writing a single line of code. A few months ago, that would’ve sounded ridiculous. Now it’s normal if the setup is good. The code wasn’t the hard part.
 The hard part was giving the agent the right tools, the right context, and an environment it could actually work in. We spent most of the hackathon on the idea, the scope, and the setup. The code came after. @MongoDB was a big part of why we could iterate that fast. I’m going to write more about the setup behind this @oscrhong @_javokhir 🚀🚀

@oscrhong Humans are bottlenecks for the agents 🙃 I think that next, more capable and more autonomous models will push usage to the new levels

For reference, the top Straude users are pushing ~$100K/year token run rate. I thought that was pushing agents to their limits. His expectations for Nvidia engineers are 2.5x global power users. There are levels to this...

Wow Jensen just set the expectation for what serious token usage looks like... "Let's say you have a software engineer or AI researcher and you pay them $500,000 a year. At the end of the year I'm gonna ask him: 'How much did you spend in tokens?' And that person said '$5,000'. I will go ape-something-else. If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed."

@AlexanderTw33ts Welcome to the club!

well that didn't take long

@ah20im Otherwise, Codex App is amazing. I used Claude Code for almost a year, and recently moved to Codex, and it feels sooo much better!

I use Codex App Server to programmatically send messages to my existing threads, that I create from Codex App. But the problem is, the conversation history in the Codex App doesn't update automatically. I have to reopen the app to see changes. It'd be nice to see auto-update or at least a manual refresh button, that updates current conversation. Or the opposite: I create a new thread with Codex App Server — I want to see it in Codex App right away, without restarting the app

What would you like to see in Codex?

@Forbiddentwo2 @claudeai No, what about you?

@markmdev @claudeai Did you try to handhold chastely with a wAIfu

A small thank you to everyone using Claude: We’re doubling usage outside our peak hours for the next two weeks.

@nika_arbnor @claudeai That's what I'm trying to figure out

@markmdev @claudeai What did you do to get suspended on claude lol?

@claudeai Cool, but it'd also be nice if someone responded to my appeal. I was a $200 Max subscriber, and didn't use it anywhere except Claude Code, and my account was suspended 2 weeks ago.

@AlessandroDuico 🤣

most people: "1" claude code users: "one"

@AlessandroDuico wooooah 🤯🤯🤯

we just crossed 100k page views

@nayaraisin Thank you! Appreciate you making it happen 🚀

@markmdev LOVE THIS! Three new influencers have dropped 🚀

We won $40,000 at hackathons without writing a single line of code. A few months ago, that would’ve sounded ridiculous. Now it’s normal if the setup is good. The code wasn’t the hard part.
 The hard part was giving the agent the right tools, the right context, and an environment it could actually work in. We spent most of the hackathon on the idea, the scope, and the setup. The code came after. @MongoDB was a big part of why we could iterate that fast. I’m going to write more about the setup behind this @oscrhong @_javokhir 🚀🚀

@AlessandroDuico Our first attempt at making social videos 🙃🙃

@markmdev You guys should shoot more commercials

@AlessandroDuico @trq212 @bcherny @claudeai I did, the problem is that they suspended my account with no reason 🙃

@markmdev @trq212 @bcherny @claudeai Did you use those tokens or did someone hack you?

Wow, this is amazing 🤓 @trq212 @bcherny @claudeai

@AlessandroDuico @trq212 @bcherny @claudeai I have no idea 😅

@markmdev @trq212 @bcherny @claudeai What happened??

@trq212 Thariq, I’m a 200$ Claude Max subscriber, I use it strictly in Claude Code and didn’t do anything wrong with it. My account was suspended today. I sent details in DM. Could you help me with resolving this please?

the team that made this are heroes, patriots

Judged the AWS x Anthropic x Datadog GenAI Hackathon in SF last Friday and I'm still thinking about it. 300+ people showed up. 83 teams submitted projects. All built in one day. The thing that got me was how far the quality bar has moved. A year ago, hackathon projects were mostly "I hooked up an LLM to a thing." Now teams are shipping full agent workflows with observability, proper error handling, the works. Multiple projects looked like they could go live tomorrow. Got to judge alongside folks from @awscloud , @datadoghq, @neo4j, @MiniMax_AI, @CopilotKit. Honestly the side conversations between judging rounds were worth the trip alone. When you get people building at different layers of the AI stack in one room, the ideas that come out of that are wild. Big shoutout to @BELLEfounders community for organizing and to AWS Builder Loft for hosting. Running a 300+ person hackathon smoothly is no joke. If you're building in the AI agent space and not going to these events, you're missing out. The energy is real.

Judged the @AITinkerers x Cactus (YC S25) x @GoogleDeepMind hackathon this Saturday. One of my favorite things about judging is you get a front-row seat to how differently people think about the same problem. The challenge was on-device AI agents. When do you keep things local, when do you route to a frontier model, and how do you handle sensitive data along the way. Teams had 5 hours and the stuff they built was wild. One team built "CloudNein," which detects PII and encrypts it before anything touches a cloud LLM. Another team basically rebuilt Siri on steroids using Gemma models to route voice commands and control apps locally. This whole space (on-device inference, smart routing, edge orchestration) is moving way faster than most people realize. The tradeoff between privacy and performance isn't theoretical anymore. People are building real solutions for it in a single afternoon. Thanks to @jacoblaes, @Henry_Ndubuaku, @nickgiometti and the whole @AITinkerers team for putting this together and having me as a judge.