@AndrewYNg Start with one agent-backed endpoint, then add logging, auth, retries, and evals before adding features. Keep the path narrow from UI to production, with one human fallback and one metric that tells you if the workflow improved.
@OfficialLoganK The same price and speed make the real opportunity workflow-specific, not benchmark-specific. Pick one recurring task, pin an acceptable latency and error budget, then compare end-to-end cost on 3.7 vs 3.8 before swapping the whole stack.
@troyaitken_ Before scaling to 100 ads, run the ICP through one real angry-customer thread or one SKU search. Tag the complaint’s job-to-be-done, urgency, and current workaround, then ask five people in that exact slice to rank the pain; repeated language is a better signal than cheap clicks.
Start with “Where’s my order?”—have the agent check carrier status, explain the delay, and offer a defined make-good (shipping refund or store credit) when the SLA is missed. Track resolution rate and time-to-resolution for two weeks; fix the human handoff if customers still need help.
If the AI doesn’t change one of these, it’s theater:
money in, money recovered, money kept, hours back that you can bill, or a skill you can still charge for.
“It answered my question” is not a metric.
Pick the number before you pick the model.
@ncodes Sketch is far superior to Figma is so many ways. It’s my go to design tool. I only use Figma when the client expects it and I hate it every time.
Opus 4.6 is useless. Paid for Pro yesterday and hit my daily limit after one prompt. Then, got back to it today to have it finish and I hit my weekly limit. Can’t do anything for 7 days. Vibe coding with cloud AIs is a joke. But running local with limited parameters isn’t an option if you need better logic and reasoning to produce a better result. It’s all a facade. Until hardware catches up, we’re stuck in limbo. But at the pace AI is moving, will we ever be able to catch up?
In the US, they made us switch to Energy Star appliances with the promise of lower electric bills. 20 years later and we’re all still waiting. It’s naive to think the elite of this world will suddenly share their wealth with those they fired (in order to lower their costs and make more profit). Lower overhead means more money for the top 0.1%. They promise to share the wealth this time only to prevent the 99.9% from fighting back. When we don’t and we’re all out of job, we’ve already lost.