Take-away? Even the best models or agents can’t replace the value of first-hand information. As a user, I still own the quality of the output

Another example: Ran into an issue migrating data to AWS—spent 30+ mins prompting ChatGPT (coz it’s supposed to have the most comprehensive knowledge on the whole internet), but left frustrated. Then it hit me—why not just read the technical docs? Turned out that the exact solution was right there!

Yeah @bsky.app should take the bots/fake accounts issues seriously.. it’s a common challenge for a high growth startup, but could become one that takes down a platform quickly if not handled properly

When can Bluesky address the fake accounts / bots issues??? There are a lot of them..

That said, let’s not forget the fundamental limitations of optimization techniques like distillation and quantization used by Deepseek, as they are inherently tied to the advancements of the original (teacher) foundation models.

Amid this trend , OpenAI will face greater pricing pressure - for example GPT4o costs $10/million output tokens vs. Llama3.3 at $0.4 and Deepseek $0.28. Hard to imagine OpenAI maintaining the same price point for devs and consumers for long