Skip to content
Back to Blog

What to do when you get stuck on a project

You received a task, tried a few solutions, and are now spinning in circles. The deadline is approaching. You think that asking for help will show you are not good enough. I have seen this fear in technical teams, from junior to senior levels. It usually causes more delays than the original doubt.

Getting stuck is not an identity. It is a state of work. The first step is to describe it in a way that someone else can see the exact same problem.

Why our brains get stuck: Herbert Simon and Karl Weick

There is a scientific explanation for the paralysis we feel when an entire system seems to fail. Computer scientist and Nobel Prize laureate in Economics Herbert Simon demonstrated in The Sciences of the Artificial (1969) the concept of bounded rationality: our working memory has limited capacity. When we try to hold all the variables of a complex architecture in our heads at the same time, we enter cognitive overload. The solution demonstrated by Simon is decomposition: dividing a large system into nearly independent subproblems and testing a single variable at a time.

In organizational psychology, Karl Weick proved this same mechanism in the classic article Small Wins (American Psychologist, 1984). When we face a blockage as a giant block ("the entire integration broke and the deadline is tomorrow"), anxiety rises and blocks logical reasoning. When we swap the huge problem for a verifiable small win, such as validating a single line of input, mental overload drops and we start moving forward again.

1. Swap "it does not work" for an observation

Write four lines before opening yet another tab:

  1. Objective: what was supposed to happen?
  2. Result: what actually happened? Include error, time, or output, without exposing sensitive data.
  3. Attempts: what did you change and what effect did each change have?
  4. Limit: what do you still not know or cannot test?

Example: "The import should finish with 100 records. It stops at record 37 with a date error. I compared the input formats and tested an isolated line; the line passed. I still do not know if there is an invalid value before it." This opens an investigation. "The import is broken" leaves everyone guessing.

2. Reduce the problem before changing the entire solution

Isolate an input, a step, or a hypothesis. If the system is large, figure out the smallest part that reproduces the failure outside the complete application: test the query directly in the database or call the isolated service. Also confirm the expected behavior with whoever requested the work. Many hours are spent perfectly solving a requirement that no one agreed upon.

Define a time limit (a timebox, such as 30 to 60 minutes) to investigate alone. The limit varies with urgency and risk; the important thing is not to let silence become a problem for the team. If you found no new evidence during this period, bring what you know to someone else.

3. Ask for help in a way that makes helping easier

A useful message can be short:

"I am trying to deliver X by Y. I expected A, but observed B. I tested C and D; C ruled out hypothesis E. My next hypothesis is F. Can you review my reasoning for 15 minutes?"

Do not hide attempts that went wrong. They prevent someone else from repeating the same path. Also, do not dump the whole problem without context. The conversation goes better when you bring a hypothesis and are willing to change it.

If there is risk to users, data, or the agreed deadline, warn whoever is responsible for the delivery early. Transparency is part of technical quality.

4. Decide what to do with what you discovered

After the conversation, choose one of the three outcomes and record the reason:

  • Continue: a testable hypothesis exists and the cost fits within the deadline.
  • Simplify: a smaller version solves the main need with less risk.
  • Pause or renegotiate: information, authorization, or time for a responsible solution is missing.

None of these outcomes are failures by themselves. The problem is leaving an important decision implicit until the last day.

5. Turn the blockage into repertoire

Upon completion, write a short note: observed signal, found cause, decision made, and what you would do earlier next time. If the error might return, add a test, alert, or documentation useful for the team. That is how I learned to value decision records and knowledge-sharing teams: they make the next person less dependent on someone else's memory.

You do not need to prove you can do everything alone. You need to learn how to investigate, communicate what you know, and take on the next step responsibly.

To continue: read How to choose a path in technology and Good decisions take work.


References and verifiable readings

Cesar Zanis / Management & careers

Good ideas keep the conversation going.

Share this article or publish the visual deck to your Stories and Feed.

WhatsApp ↗LinkedIn ↗X ↗
Cesar Zanis
Written by·Software Architect & Technology Leader

Cesar Zanis

15+ years leading software architecture, database performance, FinOps, and applied AI across enterprise operations. Built and scaled an engineering organization from zero to 20+ developers.

15+ yrs experienceArchitecture & DataCloud & FinOps0 → 20+ Devs

Enjoyed this article?

I publish practical analyses on critical systems architecture, Cloud FinOps, data, AI, and engineering leadership directly on LinkedIn.

Continue Reading

Take this analysis further with data and sources

All visual cards of this article are ready above to share on your Stories, Feed, WhatsApp, or LinkedIn with full sources.