Hey, it's me, josh - as a memoji

josh.miami

ChatGPT Replaced the Worst Part of Technical Search

Josh Echeverri
Josh Echeverri
2 min read

For years, the opening move on almost any technical problem was the same.

Open too many tabs. Search the error. Skim the docs. Click a Stack Overflow thread that is almost right, which is another way of saying wrong. Open three more tabs. Burn 20 minutes assembling enough scraps to finally have one usable thought.

That still works.

It also sucks.

What changed for me is not that some magic machine started “doing engineering.” Calm down. What changed is that ChatGPT started replacing the first stretch of dumb work that happens before the real work even begins.

That is a much smaller claim. It also happens to be true.

If I already kind of understand the shape of the problem and just need a fast first pass to narrow the space, getting a synthesized starting point is better than pretending tab-hoarding is a craft. That early orientation layer, the part where you are not solving yet, just trying to stop being lost, is where this started paying off.

That is why this mattered earlier than I expected.

The value was never “great, now I never need docs again.” The value was “great, now I do not need to waste the first 20 minutes searching badly before I can start thinking clearly.”

That is a real improvement.

It is not replacing judgment. It is not replacing experience. It is not replacing knowing when the answer is wrong. If anything, it still depends on all three. What it replaces is the slow, momentum-killing ramp from “I know this is solvable” to “I finally have enough context to start.”

That is not dramatic. It is just useful.

And useful is enough.

Right now it still feels external. I still have to bring the problem to it. I still have to know when it is bullshitting. I still have to translate whatever it gives me back into real code, real constraints, and the actual system in front of me.

So no, this is not the glorious robot future.

It is much less impressive than that.

It is just giving me back the first 20 minutes I used to waste searching badly.

That is the whole point.

And once a tool starts consistently giving you that time back, it stops feeling like a novelty pretty fast.

Related Posts

Copying Code Into a Chat Box Is Not a Workflow

The model is useful. The interface is the problem. Copying code into a detached chat box and hauling the answer back is a workaround, not a workflow.

It Is Getting Easier to Spin Up Ideas Than Finish Them

Ideation cost is down. Shipping cost is unchanged. That gap is real and I have no reason to think it closes on its own.

I Keep Opening It Before I Fully Trust It

The usefulness has outrun the comfort. That is the honest state of things right now.