It's not just that I don't understand the code, the more I use claude, I am loosing the desire to understand it even more.
While it may have gotten easy to create software, i believe maintaining it is still a constant and hard job
Silicon Valley was a documentary
WTF, $3000 Samsung TV has an always-on mic, automatic content recognition, geo tracking, network mapping, and ships your data to Samsung + 400 partners.
Bro Rooted a brand-new 2025 Samsung TV after noticing absurdly targeted ads.
Firmware was locked down, so zero-days it was.
Got a root shell.
then found: always-on voice service, ACR, geo, full telemetry pipeline, and four separate DRM systems.
This isn’t a bug, It’s the product bro.
Replying to @zebassembly
@tryramp 's agentic harness laminar also talks about the same thing in their architecture.
builders.ramp.com/post/build…
💀💀💀💀
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Cost optimisation tip: if you have your services running with python ask your agent to hill climb and rewrite it in go or rust
Started reading this book and it's quite fun. Teaches about scaling models from first principles.
(Note: it's more TPU and JAX focused, but the same principles apply to GPUs as well)
jax-ml.github.io/scaling-boo…
A lot of the companies are just going to use Claude to hill climb the optimisation road. (And if you are not you are missing out)






