DAILY ENGINEERING CHALLENGE
Agents can produce more and more code. That makes fundamentals, reasoning and technical judgment less dispensable, not more.
The Daily Challenge puts that capability in training: one runnable problem a day, from the fundamentals that let you evaluate a solution to the problems that show up once AI enters the product and agents enter the engineering process.
One challenge a day.
The goal is not to write everything without AI. It is to know enough to direct, question and confirm what is produced with it.
What you need to understand before you can judge a solution.
Data structures, algorithms, state, concurrency, invariants, idempotency and edge cases. Not because engineers have to race models, but because you cannot confirm an implementation whose behaviour you do not understand.
The problems that appear once models enter the system.
Retrieval, ranking, semantic cache, RAG, model routing, streaming, cost, quotas, observability, groundedness, personal data and prompt injection. The deterministic parts around a probabilistic system.
The engineering problems of building with agents.
Context engineering, decomposition, handoffs, budgets, execution graphs, checkpoints, tool replay, spec coverage, evals and coordination across runs.
Solutions that look right until you know where to look.
Retries that are not idempotent. Money in floating point. Broken pagination. Time zones. Unicode. Cache invalidation. Cycles. Resources that are never released. Tests that only confirm the happy path.
Here the exercise is not only to find an answer. It is to recognize why a convincing answer can still be wrong.
The rest of the page is still available. Try again in a moment.
One problem a day, with technical context and clear criteria.
Use the language, the tools and even the agents you would normally use. The point is not to prove you can work without AI.
Put the solution against verifiable cases. A plausible answer is not enough.
Ask where it breaks, what it assumed, and which trade-offs it chose.
One problem a day.
Every day you get:
Solve it with whatever tools you want, AI included.
The form is unavailable right now. Please try again in a moment.
One AI-native engineering challenge a day. You can leave whenever you want.
Do not stop when the tests pass.
Premium continues the exercise after the solution. You get:
The answer shows a solution. The review teaches you to judge it.
An implementation that passes the cases.
Why this approach works, and what it assumes.
What other path could be taken, and when it would be better.
What is gained and what is lost in the choice made.
The plausible implementation that fails, and the exact reason.
Fundamentals used to help you produce the code. Now they also help you judge code produced by others, human or agent.
You cannot confirm what you cannot evaluate.
One problem a day. Not to prove you can code without AI, but to make sure you can keep doing engineering with it.
Start with one problem a day, for free. When you want to turn each problem into a full engineering review, move to Premium.
Already getting the challenge? Come back tomorrow.