C-TIER news
Kimi K2 in an agentic coding class — open models now viable for web apps
100k ft — What is this?
Kimi K2 in an agentic coding class — open models now viable for web apps
50k ft — Why care?
Liz the Developer reports her agentic software-development class switched to open-source models (Kimi K2) after Claude token cuts, and the class still shipped a 'Minecraft in space' web game. Light on detail — mostly a vibe check that open models can keep up for web-app builds.
10k ft — Key takeaways
• Open-weight models like Kimi K2 can replace a paid Claude subscription for classroom-scale web-app builds
5k ft — Tools & techniques
Anthropic, Claude, Kimi
AI agents, agentic software development
1k ft — Deep-dive / try it
Big news: Kimi-k2 did keep up with Claude for web apps
Transcript excerpt
Anthropic is really committed to open source lately, you guys. First, they open source clod code. Second, they cut us all down to one quarter of the tokens available. And so today, I used a bunch of open source models in my agentic SDLC class. So now you can use an open source model in my agentic SDLC class. Previously, this required a $200 subscription. So everybody that I told you needed to come back next time when like an open source model was ready. We're ready. We made Minecraft in space. It's Minecraft in space. The thing is when you can summon any kind of software that you want from the ether, like you can't avoid making your own campus. Like you just can't pay for anymore. And this is not animal crossing because I do not own the IP to animal crossing. And just because you can fish doesn't mean that it's animal crossing. It's not. It's different. It's different. Oh, hey, I've been looking for one of those. Okay. So it was a good class. I'll see you in the next one.
More from @lizthedeveloper
- B-TIER Game decompilation with AI — which reconstruction methods stay legal
- C-TIER Data reclamation game — gamifying opt-outs from data brokers
- S-TIER Stay-secure privacy checklist — de-dox yourself from AI and data brokers
- C-TIER Agentic misalignment benchmark — models now know they're being tested
- C-TIER METR rogue-deployment study — agents resisting shutdown, rant edition