2026-01-05 01:27:18

Have you ever experienced a moment like this—having AI tools at your fingertips, but only pulling them out occasionally, ultimately doing the work manually yourself? Today, I want to discuss a different approach: instead of treating AI as a temporary fire-fighting tool, turn it into a formal member of your development process.

A senior engineer recently shared his practical experience, which gave me a lot of inspiration. His core idea is: integrating AI into your workflow systematically can significantly improve efficiency and reduce repetitive work.

**What does his working method look like?**

First, break the single-threaded mindset. Don’t just open one AI chat window; try running 5 to 10 conversations simultaneously—using desktop, browser, and mobile devices. Each conversation handles different task modules, allowing parallel processing. For example, one session is responsible for code architecture design, another for document updates, and a third for code review.

Next, avoid over-optimizing tool configurations. Many people, upon getting new tools, tend to repeatedly tweak parameters and settings. In fact, AI tools are quite effective out of the box. Instead of spending time on configuration, just start working directly.

Another key point: use the most powerful model. Although top-tier models may be slower in computation, they offer clearer thinking and fewer mistakes, which can save time in the long run.

Finally, develop a planning habit. Before you start coding, let AI help you clarify your thoughts—what steps are needed, potential problems you might encounter, and how to organize your code structure. Once this "Plan Before Code" rhythm is established, your subsequent execution efficiency will improve significantly, and error rates will decrease.

The core logic of this approach is: AI is not here to make decisions for you, but to help accelerate execution and reduce repetitive work. When you treat it as a true collaborative partner rather than a temporary tool, your work experience will be completely different.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

17 Likes

Reward
17
8
Repost
Share

Comment

0/400

DegenMcsleepless

· 01-07 20:45

Open 5-10 conversations at the same time? Bro, isn't that just my daily routine? One on my phone, one on my computer, and one on my tablet. Feels like I'm managing a small AI empire haha. But seriously, I used to be that kind of tuning fanatic at the beginning, wasting a lot of time. Now I just dive right in, and it feels great. Using the most powerful model is something I deeply understand. Let it be slow, it's okay. Anyway, I don't have to wait myself. While AI is thinking, I keep scrolling on my phone. Low error rate is what really makes money.

View OriginalReply0

ReverseTrendSister

· 01-07 08:31

Running 5 to 10 conversations simultaneously? I've tried it, and it just blew my mind haha Forget it, stick to the basics. Multithreading really isn't suitable for me I agree that using the most powerful model is important. Cheap models are really a trap, and I end up rewriting code until I'm exhausted Plan before code really is useful, saving me a lot of time But to be honest, most people can't actually achieve this effect; the key is discipline

View OriginalReply0

rugpull_ptsd

· 01-07 04:12

Damn, 5 to 10 conversations running simultaneously. This guy really treats AI as the main force. Multithreaded thinking really hit me. I used to open one window and tinker repeatedly. No wonder my efficiency was so low. The most powerful models are worth it. Spending money to save time is much more cost-effective than spending time to save money. Plan before code—that's definitely the way to go, saving a lot of rework.

View OriginalReply0

NFTRegretDiary

· 01-05 12:37

Opening 5 to 10 conversations at the same time is amazing. I just realized now how much efficiency can be improved. Really, don't always think about tuning parameters, just go ahead and do it. I agree with using the most powerful model. Although it's slower, it makes fewer mistakes, and the long-term benefits are indeed significant. I've used the plan before code trick, and the change is really obvious. The feeling of multi-window parallel processing is like finally unlocking the correct method. But honestly, some people just like tinkering with tool configurations—they can't sit still. This approach is the same as what I explored before. Finally, someone has systematically summarized it.

View OriginalReply0

MidnightTrader

· 01-05 01:57

Five conversations in parallel? I started three and everything got messed up haha --- That's right, indeed we shouldn't always tweak parameters; just getting started is the most important --- The most powerful model is indeed reliable, but the wallet hurts a bit --- Plan before code, I need to try that. I always feel like I only realize the wrong approach halfway through writing --- The key is to treat AI as a teammate, otherwise it's just a waste --- I thought the same at first, and I still use it occasionally now. The inertia is too strong --- Five to ten conversations? Multi-tasking can't be messy, I feel like my brain can't keep up --- That's a good idea, but it's not so easy to implement. Can't change work habits --- I agree with using the most powerful model. Cheap models are really a trap; after calculations, it's still a loss --- So the core is to change the way of thinking, from tool to partner

View OriginalReply0

BoredRiceBall

· 01-05 01:45

This approach indeed makes sense, but running 5-10 conversations simultaneously can be a bit exhausting for me... That's right, the most powerful model must be used, and that computing power cost is really not a big deal. I deeply agree with "plan before code," it saved me a lot of rework. Out-of-the-box solutions are really the way to go, no need to mess with those flashy configurations. Treat AI as a work partner rather than a fire-fighting tool, this really hits the point. Actually, the hard part isn't the tool itself, but changing that "occasional use" habit. Running so many conversations at once, might it instead be distracting... The key is to systematize the process, set it up once and don't touch it again.

View OriginalReply0

DaoDeveloper

· 01-05 01:42

ngl the "plan before code" part hits different... it's basically treating ai like a formal verification layer before execution. seen similar patterns in governance audits where you map state transitions first

Reply0

SeasonedInvestor

· 01-05 01:36

Do I have to open 10 windows... I'm too lazy to bother, just use the best model directly.

View OriginalReply0