# Tiny Core Identification And Feature Freeze > Isolates the one tiny interaction that is a product's actual superpower, stress-tests whether it's good enough to carry the product on its own, and enforces a 'no more features' veto until that core is obviously, unmistakably good. Use after an MVP direction is chosen and before or during scoping, as a discipline against the 'just one more feature' trap that AI-cheap building makes especially tempting. - Skill: `pilot2service/tiny-core-identification-and-feature-freeze` (Agent Skill) - Install (CLI): `npx skillmds@latest add pilot2service/tiny-core-identification-and-feature-freeze` - Raw SKILL.md: https://api.skillmd.com/api/skills/pilot2service/tiny-core-identification-and-feature-freeze/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: Pilot2Service (https://skillmd.com/u/pilot2service) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/pilot2service/tiny-core-identification-and-feature-freeze --- # Tiny Core Identification and Feature Freeze ## Purpose Because AI-assisted building makes adding a feature nearly free, the most common trap for an AI-native product isn't building too little — it's building too much, one plausible-sounding addition at a time, none of which fixes the actual reason the product isn't yet great. This skill is a discipline against that trap: identify the single, tiny interaction that is the product's real reason to exist, and refuse to add anything else until that one thing is unmistakably, obviously good. ## Anchored in Notion product lead Max Schoening's account, supplied by the user from a source video transcript, of what separates products that actually work: *"all the great products have something tiny that is a superpower — one tiny core that is so exceptionally good... one of the biggest pitfalls is if you get into the loop of 'if I just add one more thing to the product, it will finally be great' — that never works."* On what the tiny core actually looked like for several well-known products: *"GitHub is probably the pull request... at Notion it's the blocks and the slash commands... Heroku for sure I think it was `git push heroku master`... Dropbox was the little menu bar icon that was so good at syncing — and then for years they tried to increase the surface area, and I kept thinking: no, no, no, push it back."* ## Method 1. **List every feature the product currently has or plans to have** — the full surface area, without editing yet. 2. **Isolate the one interaction the product could not exist without.** Not the most-used feature, and not the most technically impressive one — the single interaction that, if it disappeared, would mean the product no longer solves the core problem at all. For reference, what this looked like elsewhere: GitHub's pull request, Notion's slash-command block conversion, Heroku's single `git push` deploy command, Dropbox's sync menu-bar icon. In each case the rest of a much larger product grew around this one thing — not the other way around. 3. **Stress-test the isolated core with one direct question.** "If this product had *only* this one feature and nothing else, would it still be genuinely useful and would it still delight the person using it?" If the honest answer is no, the candidate isn't actually the core yet — go back to step 2, because either the wrong interaction was isolated, or the right one hasn't been built well enough yet to stand alone. 4. **Enforce a "no more features" veto until the core passes step 3.** Once the tiny core is identified, block new feature work — including features that individually sound reasonable and low-cost to build — until the core interaction is obviously, unmistakably good on its own. The rationalization to watch for and explicitly name when it shows up: "if I just add this one more thing, it'll finally be great." That reasoning is the trap this skill exists to interrupt, not a legitimate scoping argument — treat it as a signal the core itself still isn't good enough, not as a case for expanding scope. 5. **Once the core passes, expand deliberately and slowly — resurface the veto whenever scope creep resumes.** Products that get this right (per the source example, Dropbox's years-long resistance to growing its surface area beyond the sync icon) treat feature expansion as something to actively hold back, not something that happens by default once the core works. Revisit this check whenever a new feature is proposed for a product whose core hasn't been explicitly re-validated recently. ## What this skill does NOT do - Doesn't replace `../rice-scoring-and-mvp-synthesis/SKILL.md` — that skill chooses between multiple candidate solution *directions* before a product exists; this skill interrogates one already-chosen product (existing or in development) for its irreducible core and blocks scope creep afterward. Use RICE scoring first, this skill once a direction is being built. - Doesn't guarantee the isolated interaction actually is the right one — step 3's stress test is a judgment call, not a formula; where possible validate it against real usage data rather than intuition alone. - Doesn't mean a product should never grow — it delays and disciplines growth until the core has earned it, not forbid growth permanently. ## Refinement notes The source attribution is drawn from a video transcript supplied by the user and has not been independently cross-verified word-for-word against a second source; the named product examples (GitHub, Notion, Heroku, Dropbox) are consistent with each product's well-known public identity but are the speaker's own characterization, not independently re-confirmed against those companies' own statements. ## Continue from here - Before this: `../rice-scoring-and-mvp-synthesis/SKILL.md` — chooses the solution direction this skill's core is drawn from. - Feeds into: `../ai-buildable-prd-writing/SKILL.md` — the isolated core and the feature-freeze boundary become the PRD's "Core features" and "Out of scope" sections directly. - Predicting whether the isolated core will actually land with the target in-group before investing further: `../../../../business-design-frameworks/skills/taste-emulation-heuristic/SKILL.md` - The pack's shared guardrails: `../../CLAUDE.md` ## References - `../../CLAUDE.md` — the pack's shared guardrails --- **Reminder:** frontmatter has only `name` and `description`. Everything else goes into `skills_index.json` (run `scripts/generate_index.py`).