A reader asked a great question:
AI is evolving rapidly, with constant iteration, and new concepts/technologies/products emerge endlessly. However, my current workload is heavy (the work itself isn’t directly related to AI), so even for many popular AI-related things, it’s impossible to spend time understanding/learning/using them deeply. On one hand, I want to keep up with the AI wave; on the other hand, my current energy doesn’t allow it. What principles can help make the following judgments: “Which things don’t require time? / Which things just need a basic understanding? / Which things require some time to learn? / Which things are necessary for deep usage and learning?” and why these principles exist.
This question is very typical. I’m the same, making choices while processing a massive amount of information daily. Ultimately, it’s about limited time, AI new things coming out too fast, and the fear of missing out (FOMO).
You can try using a four-quadrant method to filter.
To judge whether a new AI thing is worth your time, ask two questions:
First: How close is it to my current productivity?
Note, not how close to the “AI industry,” but to your productivity. Can it save you time or improve the quality of output for the things you do daily, the tasks you repeat? If you can name specific scenarios, it’s close. If you think for a long time and can’t say how it helps you, it’s far.
Second: How long is the shelf life of this knowledge?
Some things you learn today will still be useful three years later; the time invested generates compound interest. Some things you learn today might have their names changed in three months; learning them becomes a sunk cost.
With these two axes intersecting, the decision logic for the four quadrants becomes clear.
The horizontal axis determines “whether to learn” — the closer to your current productivity, the more worth investing time.The vertical axis determines “whether what you learn will expire” — the longer the shelf life, the more lasting the return on investment.
The top-right corner (Close + Long) is the most scarce and most worthwhile, usually only one or two at a time.The bottom-left corner (Far + Short) is the noisiest area, filled with funding news and hype; filter directly.
The diagonal line in the middle is the main direction for time allocation: from bottom-left to top-right, investment increases, quantity decreases.
Bottom-Left: Skip Directly
Far from productivity, short shelf life.
This quadrant has the densest noise. Funding announcements, preprint papers, strategic moves by big companies all pile up here. The characteristic is: after reading the introduction, you can’t clearly say who it helps or what it does.
For example:
-
• Monthly changing model benchmark rankings -
• AI startup funding news -
• Internal drama gossip at model companies -
• AI chip parameter details (unless you work in hardware) -
• Various AI wrapper products (most won’t survive half a year)
Skipping causes no loss. If it’s truly important, it will still be around in three months; you can look then.The attrition rate in the AI field is extremely high; investing time too early incurs the greatest sunk cost.
Top-Left: Maintain a Mental Map
Far from productivity, but long shelf life.
These things have become industry common language; you can’t converse without knowing them, but you don’t need to use them.
A few examples.RAG (Retrieval-Augmented Generation), now used in almost all enterprise AI applications; chatbots citing internal company documents rely on this. You don’t need to build a RAG pipeline yourself, but you should know “it makes AI retrieve relevant information before answering, instead of purely fabricating from memory,” so you can keep up when colleagues mention it.
Similar ones include:
-
• Chain-of-Thought: Why reasoning models like o1/o3 “think for a while” before answering -
• Scaling Laws: Why bigger models are smarter, why training costs are astronomical -
• AI Hallucination: The principle behind AI confidently making things up -
• Multimodality: How text, images, audio, video are fused into one model
These concepts have long shelf lives, and you’ll encounter them repeatedly in daily news, colleague chats, and product introductions.
The goal for this quadrant is to maintain an “AI mental map”. You don’t need to visit every city on the map, but you should know where they are and roughly what they’re like.
Method: Read one good explanatory article, don’t get hands-on, don’t set up environments, finish in 15 minutes.
Bottom-Right: Worth Trying Hands-On, But Don’t Invest Too Much Effort
Close to productivity, but shelf life uncertain.
Things in this quadrant: you can say specifically how they help you, but you’re unsure how long they’ll stay popular or if their form will change drastically.
AI image generation tools are a classic example. Making PPTs needs images, writing articles needs header images, social media needs visual materials — almost all knowledge workers can use them.
<p style="margin: 1.5em 8px;letter-spacing: 0.1em;color: #3f3f