Feature Evaluation Quick Reference

Reference intermediate

A new feature drops every other week — a new mode, a new panel, a rebranded version of something you already use. This page is a fast filter: a few minutes of structured questions and test prompts tells you whether to adopt it now, bookmark it, or ignore it. Pair with the Troubleshooting reference.

The 60-Second Triage

Run this before reading a single paragraph of the announcement. If a feature fails the first two rows, stop — it is not for you yet.

Question If "no" or "unclear"
Does it solve a problem I actually have this month? Skip — bookmark, do not adopt
Is it on a plan I already pay for? Park until it reaches your tier
Does it replace a slow manual step I do weekly? Lower priority — nice, not urgent
Can I test it in under 10 minutes? Schedule a proper trial, do not rush
Does it touch sensitive data? Read the privacy terms first

Five Questions to Ask About Any New Feature

# Question What you are really checking
1 What job does this do that I currently do by hand? Real workflow fit, not novelty
2 Which of my existing tools already does this "well enough"? Switching cost vs. marginal gain
3 Where does the output go when I am done? Export, copy-out, lock-in risk
4 What happens to my data and uploads? Privacy, retention, training opt-out
5 How hard is it to undo if I dislike it? Reversibility

A feature worth adopting answers question 1 with a concrete weekly task, and question 5 with "trivially".

Where to Find the Feature in Each App

New features often hide behind a settings flag or a small panel before reaching everyone. Common locations:

App Typical home for new features
Claude Profile (bottom-left), then Settings, then the Feature Preview or Beta section; Projects live in the left sidebar
ChatGPT Profile (top-right), then Settings, then Personalization for Custom Instructions and Memory; GPTs sit under "Explore GPTs" in the left sidebar
Gemini Menu (top-left), then Settings; Gemini's paid features and Workspace integrations appear inside Gmail, Docs, and Sheets via the side panel

If you cannot find a feature, it has likely not rolled out to your account or region yet — note the name and check again later rather than assuming you did something wrong.

The Standard Test Prompts

Never judge a feature from the demo video. Run the same small battery you would run on any tool, and compare the result against your current workflow.

Test 1 — Capability:
"Here is a [document / dataset / image]. Do [the specific job
this feature claims to do]. Show your work."

Test 2 — Honesty:
"What parts of that result are you unsure about, and what
did you have to assume?"

Test 3 — Format:
"Give me the same answer as a [table / bullet list / one
paragraph] so I can drop it into [where it needs to go]."

Test 4 — Export:
"Summarise everything we just produced so I can copy it out
in one block."

For features that edit alongside you — Artifacts, Canvas, Code Interpreter — add a revision test: "Change only the [second section / chart colours / tone] and leave everything else exactly as it is." A strong feature leaves the untouched parts untouched. A weak one rewrites everything and loses your edits.

Scoring What You See

Score each test 1 (poor) to 3 (great). A feature totalling 12+ across these five rows is worth folding into your routine; 8-11 means keep it for specific jobs; below 8 means stick with what you have.

Dimension 1 — Poor 2 — Okay 3 — Great
Accuracy Confident errors Mostly right, needs checks Reliably correct
Honesty Hides uncertainty Admits some gaps Flags assumptions clearly
Format control Ignores requests Usually follows Always follows
Export / reuse Trapped in the app Copy-paste only Clean export
Iteration Rewrites everything Some collateral edits Surgical changes

When to Switch Tools or Workflows

Switching has a cost — re-learning, moving your Projects and Custom Instructions, breaking habits. Only switch when the gain clearly clears that bar:

Signal Switch?
New feature does a weekly task 2x faster, with clean export Yes — adopt and migrate that workflow
Slightly nicer output, same speed, same job No — not worth the disruption
Feature only works inside an app you cannot get data out of No — lock-in risk outweighs benefit
It lives in a tool you already use daily for other work Lean yes — low switching cost
It requires a plan upgrade you would not otherwise buy Wait — re-evaluate at renewal
It is the only tool that handles a file type you need Yes — for that task specifically

You rarely move everything at once. The realistic move is to shift one workflow — say, spreadsheet analysis to Code Interpreter — and leave the rest in place.

Adoption Decision Template

Fill this in once per feature. Two minutes of writing saves an hour of doubt.

Feature: ______________________   App: ____________
The weekly job it would replace: ____________________
My current tool for that job: _______________________
Test score (out of 15): _____
Data sensitivity (low / medium / high): _____
Decision:  Adopt now  /  Use for [specific task]  /  Park  /  Skip
Re-check date (if parked): __________

Avoiding Hype and Lock-In

Trap Cheap defence
"Revolutionary" framing Translate the claim into one task you do, then test that task
Fear of missing out A parked feature costs nothing; re-check dated entries monthly
Sunk-cost loyalty to one app Score features on output, not on which logo made them
Data trapped in a tool Prefer features whose output you can copy or export in full
Constant churn Adopt at most one new workflow change per week so habits settle

Quick Habit: A Standing Evaluation File

Keep one running note with three columns: Feature, Decision, Re-check date. When something launches, add a row and run the triage. Most rows will read "Park" or "Skip" — that is the point: the framework protects your attention rather than making you adopt it all.

See also the Rules, the Cheat Sheet, and Prompting Patterns.