AI-assisted engineeringApplied
Agent evaluation
Easynot started
Mediumnot started
Hardnot started
StatusFormatCompany style
Easy
Decide which log anomalies page the on-callAgent build · Google styleGoogle styleAgent buildGoogle style15 minDecide when a summary agent should stay quietAgent build · Microsoft styleMicrosoft styleAgent buildMicrosoft style10 minKeep the generated tests that catch real bugsAgent build · Meta styleMeta styleAgent buildMeta style10 minTriage CVE advisories with a bar per severityAgent build · Microsoft styleMicrosoft styleAgent buildMicrosoft style12 minFind open items the handoff summary droppedAgent buildAgent build10 minRoute accessibility findings without overclaiming a clean scanAgent build · Microsoft styleMicrosoft styleAgent buildMicrosoft style10 minTell a real benchmark regression from noiseAgent build · Google styleGoogle styleAgent buildGoogle style10 minDecide when a support agent may answer aloneAgent build · Salesforce styleSalesforce styleAgent buildSalesforce style10 minPick the tests to run for a diffAgent build · Meta styleMeta styleAgent buildMeta style10 minFill a labeling queue without starving the rare classAgent build · OpenAI styleOpenAI styleAgent buildOpenAI style12 minPick the lowest safe threshold for duplicate suggestionsAgent buildAgent build10 minScore whether a capacity forecast's ranges hold upAgent build · Google styleGoogle styleAgent buildGoogle style10 minMedium
The agent looped and gave up: find the first bad spanDebug the agent traceDebug the agent trace30 minScore an extraction agent's amounts within a toleranceAgent buildAgent build25 minScore a retriever's ranking when its scores tieAgent buildAgent build25 minThe order agent quoted a tracking status three days staleDebug the agent traceDebug the agent trace30 minThe invoice agent retried a file the OCR service would never acceptDebug the agent traceDebug the agent trace30 minHard
Spec the flag SDK every service calls per requestVibe codingVibe coding25 minRank agents on a rubric with partial credit and tiesAgent buildAgent build40 minScore an agent's majority vote over k sampled answersAgent buildAgent build40 minRank agent versions against the fewest judge verdictsAgent buildAgent build40 minThe review bot read a finished check run as a passing oneDebug the agent traceDebug the agent trace45 min