Mikito未来辉

紐約時報找了台北的3流砖家

8) AGI 2.0 should abandon the “replace humans” fantasy. Let the robots take over. It’s not necessarily bad — machines only need to understand binary. All the UI, interactions, and pretty facades are redundant.

7) If there’s no fundamental infrastructure progress, only model scaling, then humans will indeed be replaced — especially by the Anthropic-style vision. We’re still stuck in Stage 1, and that goal can never truly be completed.

6) Is it Anthropic’s version — “human-like replacement” leading to mass unemployment, aligned with human values? Or OpenAI’s version — self-inventing systems that increasingly resemble backend engineering?

5) Frontend has basically run out of real technical progress. From now on it’s just agents going wild — carving fancy patterns on the car door and then welding it shut. The definition of AGI remains extremely vague.

4) Back in the day, React had over a dozen language bindings. It was truly primitive — slash-and-burn development with no common language. This is not first-principles engineering like Redis, where everything is built layer by layer in a solid, logical way.

3) AI is better suited for open-source products rather than low-level infrastructure. But at the product layer, programmers are losing their voice. Management is self-interested and performance-driven. It’s inherently competitive.

2) Meta invented React for universality and cross-platform use. Ironically, their own models still can’t fully achieve that. On top of that, they built own language &eco.The beautiful old vision of SaaS companies contributing to open source is going to be shattered by agents.

1) Anthropic upgraded their model output from Markdown to HTML. In the future it’ll look prettier and be more universal, but the code will become extremely fragile.

4/ manage agent implementation plan with ⁠/artifacts⁠. You can review and approve generated plans before they run. if things look off, just leave a comment right on the artifact to steer the agent back on track.

> Grab a Grok subscription > Proceed to auth w/ "hermes model" > Turn on x_search on "hermes tools" > Search real-time news on X with x_search > Combines with Deep Research Pipeline + StableEnrich (agentcash) + Cookie MCP + Hindsight for high quality research on any topic ​ Result: Top-tier alpha at your doorstep everyday paying $40-50 a month (delivered by a self-improving agent that gets better everyday)