Throwing tokens at my problems

That Little Snitch red warning had me 😱

Anthropic models cost a magnitude (we are talking 20x plus) more than DeepSeek V4 Flash or z.AI’s GLM 5.3 Flash for the same tokens. Most workloads don’t need Fable let alone Opus levels of intelligence, and these will refuse to help you when and where you truly need them like with security. The 5-hour and weekly limits on Pro and Max plans will limit your ability to ship or cause you to overpay putting a cap on your efficiency. Eventually you’ll be forced to use their API rates which are the most expensive. What most people miss is a Software Factory, a repeatable workflow that yields good enough reliable software you can spec, build, and ship that sizes and break downs the work into workable units. You shouldn’t be building software but the machinery that does it for you.

They say “The government has a debt problem” I hear “The government is financially irresponsible and also broke”

So now I’m moving from having a single agent that does everything to a primary I talk to that controls multiple ephemeral or static agents with similar capability as the primary. All comms happen over NATS so I don’t have to bother with “discovery” across hosts.

What’s hot on Nostr right now?

Lake Epstein sounds better than Lake America

🏡🚗🛣️⛺️🪵🔥🥩🍻🌞🌚🔭🚗🏡

Agent harnesses are great. But you know what’s even more cool? Building your own. Ok to build on top of an existing one, but make it yours.

I hadn’t heard of https://nostrhub.io/apps Sweet.