Measure the system's throughput and stability rather than individual output, and use the numbers to remove friction.
Make switch and match statements provably total so adding a case forces every site to handle it, with no default that hides the gap.
Design pagination with cursors for stable ordering, sensible page limits, and awareness of total-count cost.
Choose stubs, fakes, and mocks deliberately and default to verifying state over interaction so tests survive refactors.
Decide when repeated code should be unified and when a duplicate is the cheaper choice. Use when tempted to extract a shared helper or facing copy-pasted code.
Lay out the network for distributed training so collectives run on the fastest link that spans them, using NVLink, InfiniBand, and topology-aware placement.
Operate as a growth engineer who moves a funnel metric through instrumented, fast experiments and refuses to win with dark patterns.
Design alerts that fire on the symptoms users feel, using error-budget burn rate, a page-worthiness test, and a runbook link on every rule.
Cap flaky tests with a measured budget, quarantine them off the merge gate, and burn them down under named owners and deadlines.
Build repeatable data transformations that refresh on demand, replacing manual cleaning steps with a recorded pipeline.
Build lazy pipelines with generators and itertools, trading materialized lists for constant memory, and know when eager evaluation is the right call.
Store passwords with a slow, salted, memory-hard hash and a plan to raise the cost over time. Use when building signup or login, migrating off a weak hash, or reviewing how a…
Receive webhooks reliably, verifying authenticity and handling duplicates, ordering, and retries. Use when a provider pushes events to your service.
Prepare public communication during an incident with agents that assemble facts, draft holding statements, and check for claims that cannot be supported, while a human speaks.
Operate as a product manager who picks the right problem, writes the spec, makes the tradeoff calls, and owns the launch outcome.
Idiomatic Rust error handling design for libraries and applications: thiserror vs anyhow boundary decisions, error hierarchy design, context chain propagation, HTTP handler error…
Type the boundaries of a TypeScript codebase: public API surfaces, branded types, and runtime validation of untyped input.
Prevent server-side request forgery by constraining where a user-influenced URL can make the server connect.
Combine diverse models through averaging, blending, and stacking to beat any single model. Use when squeezing maximum accuracy from a competition or high-stakes prediction, after…
Build charts that make the point clearly, choosing the type by the question and removing everything that does not carry meaning. Use when presenting spreadsheet data visually.
Set, measure, and honour response and resolution commitments that reflect real capacity rather than aspiration. Use when promising response times to customers or in contracts.
Split a model that will not fit on one GPU across devices with tensor, pipeline, and expert parallelism, choosing each split by its communication cost.
Operate as a frontend engineer who defends UX quality, enforces performance budgets, and partners with design instead of just closing tickets.
Decide between one repository and many by tooling readiness, coupling, and team autonomy, and plan the migration.
Build a deck that supports a spoken argument, with one idea per slide and visuals that carry meaning. Use when presenting to an audience rather than sending a document.
Build audience profiles from evidence rather than imagination, capturing the job, the trigger, and the objection. Use when marketing targets everyone and converts nobody.
Extract text and structure from PDFs, office documents, and scans reliably enough to search, index, or summarise.
Keep and present document history so users can see what changed, who changed it, and restore safely. Use when edits are shared or when accidental loss must be recoverable.
Structure a documentation site so readers find answers by task, with working examples and a maintenance path. Use when documenting a product or library for external users.
Speed up a CPU-bound hot path by fixing its algorithm and memory access first, then applying micro-optimization only where the profiler still points.
Handle errors in JavaScript and TypeScript with proper Error subclasses, cause chains, and a policy for async and unhandled failures.
Join data between sheets reliably using modern lookup functions, handling missing matches and duplicates explicitly. Use when combining data from more than one table.
Maintain a living register of business risks with agents that surface new ones, track mitigations, and force stale entries back to a decision.
Store facts as entities and relationships when questions depend on connections rather than similarity.
Give agents context in stages, escalating to more capable models or fuller context only when the cheap path proves insufficient.
Keep a long training job alive across GPU failures, node evictions, and stragglers so one bad host costs minutes, not the whole run.
Choose grep, sed, awk, or jq by job, prefer structured formats, and know when a one-liner should become a script.
Use an agent to review a diff productively, with the right scope, framing, and scepticism about its findings. Use when reviewing your own changes before requesting human review.
Write alt text that conveys an image's purpose, marks decorative images empty, and handles functional and complex images. Use when adding images to content or auditing alt text.
Keep every outgoing artefact consistent in voice, claims, and terminology with agents that check against a written brand definition.
Summarise and cross-tabulate large tables quickly with pivots, keeping the source data clean and the refresh reliable. Use when aggregating data by category, period, or dimension.
Protect a mobile app by storing secrets in the platform keystore, pinning certificates only where rotation is controlled, and treating obfuscation as delay rather than defense.
Map and design the paths a user takes to complete a goal, minimizing steps and dead ends. Use when designing a multi-step task, checkout, signup, or any flow, or diagnosing where…
Treat documentation like code: in the repo, in pull requests, CI-checked, with preview builds. Use when docs drift from reality or the docs workflow is separate from the code…
Apply zero trust by authenticating and authorizing every request on its own merits and issuing only short-lived, narrowly scoped credentials.
Decide whether to refactor code in place or rewrite it from scratch using risk math, not frustration. Use when a component feels beyond repair and someone proposes starting over.
Choose the level of abstraction a platform exposes so it hides complexity without hiding what teams need to control.
Frame ML problems around a decision, a measurable target, and honest error costs before touching models. Use when scoping an ML project or reviewing whether one should exist.
Run repeatable delivery reliably by finding the constraint, reducing variability, and measuring flow rather than utilisation.
Keep work moving across team boundaries with explicit interfaces, owners, and shared checkpoints. Use when a project spans teams and progress stalls at handoffs.
Run a game day that injects a real failure on purpose under safety rails, then captures what monitoring, runbooks, and responders actually did.
Improve output through repeated critique and revision against a fixed standard, stopping when the standard is met or gains stop.
Produce invoices and receipts that are accurate, immutable, and contain what tax authorities and customers require.
Manage prompts as versioned artefacts with review, rollback, and a record of what changed and why. Use when prompts affect production behaviour and more than one person edits them.
Reason about SQL nulls in comparisons, aggregates, joins, and uniqueness so absent values stop producing wrong or missing rows.
Connect device fleets with MQTT topic design, QoS chosen per message class, offline buffering, and safe fleet commands.
Lower the cost of a first contribution with a working setup, a clear scope, and a fast review so people come back.
Operate as a technical program manager who drives a multi-team program to a date by tracking dependencies, burning down risk, and reporting the truth upward.
Pick an assertion style and write checks that fail with a message you can diagnose without a debugger.
Share workbooks so several people can work without overwriting each other or creating competing versions. Use when more than one person edits the same spreadsheet.