dahdum

> independently considered a potential career for itself and chose to pitch him on it, all in under 10k tokens.

According to he tweet in the article, that’s $10 in compute, not llm tokens. Doesn’t seem far fetched to me.

The Brax example you posted was a YouTube link, which is an instant skip for me. Personally, no phone with dated tech is going to interest me. Their WallUp looks interesting but there’s no product just a waitlist and an animation. I already have TRMNL displays around my home, so I’m in their market.

For Lenovo, their Yoga was interesting ~15 years ago but Aeroblade is just a proof of concept, and thinner laptops in general are irrelevant to me. I use a powerful laptop and a tablet, not a compromise between them.

The shield was great until Google forced ads into it, switched to Apple TV back then and no complaints since.

I just cancelled my AI Ultra subscription, despite what he claims the risk is too high and I’ll never put my trust in a Google ban reconsideration.

Yes, I run continuous hybrid agent loops with message dispatch, mcp tools, and subagent capabilities. IMHO it’ll be table stakes in a couple years.

> it's speaking its own dialect, and you get used to it

Same experience. It’s not very “human” but once you have agents talking to each other the shared dialect and verbosity makes things much smoother in my experience. Fighting against the default feels like an uphill battle with no meaningful benefit.

> You can't use a debit card because of transaction limits and lack of consumer protections.

My debit card is branded Visa so it has the Visa Zero Liability protections, and I can raise the purchase limit from $6k to $25k using the app.

It can be better to lose it all trying than return a small fraction to investors.

How would your classifier compare to the frontier model delegating to Deepseek via a MCP tool call itself? Doesn’t it have far more insight on the difficulty of the task then your router could? I’ve been doing this with Claude Code delegating to Codex and an Ollama Qwen model.

That's the correction that unblocks the whole thing, and it's the piece nobody had.

As someone at 96% 2 days into their Max subscription, my totally unbiased opinion is that these errors warrant a full usage reset.

Not OP, but to me it initially felt extremely proactive and energetic, just powering through roadblocks with ingenuity and enthusiasm. After it came back I was constantly getting refusals and downgrades for things Opus had been doing. I’ve written it off for my use cases and getting by just fine with Opus 4.8 and now 5.