Drive-Thru Macros: pick your diet and a chain, get the best orders with the exact modifications

Source: the discussion ~/vault_fbs/conversations/2026-09-19-fast-food-macro-finder.md, ratified 2026-09-19. Written by the director on the native line ([[M-074]], FACTORY § 9a); it replaces the line's first draft, which planned a web product. The v1 chain list was added 10:50 CT from research/chains-2026-09-19.md. Competitive figures are from US App Store listings fetched 2026-09-19 through a summarizer and are to be spot-checked before they are quoted anywhere public.

Problem. Someone on a diet pulls into a drive-through with two minutes to decide. The chain's nutrition PDF runs a dozen pages, and a logging app tells them what they ate after the fact. They want the order itself, in words they can say at the speaker ("double patty, no bun, no ketchup"), ranked for the diet they are actually on, with numbers they can trust.

Who. An adult on a named, macro-driven diet (high protein, low carb, keto, Atkins, carnivore, Mediterranean, calorie cap) who eats fast food weekly or more and owns an iPhone. Wayne is the first user. Consumer only; no institutional buyer.

Must.

• Core loop. Pick a diet, pick a chain, get the top ranked orders. Each shows protein, carbs, fat and calories and the modifications as words to say at the counter.
• Named diets as profiles. A diet is a data file: hard limits plus a ranking function. v1 ships high protein, low carb, keto, Atkins (one profile per phase), carnivore, Mediterranean and calorie cap. A per-meal target is set by hand. Adding a diet adds a file and no code.
• Deterministic and offline. Scoring is pure Swift in a Core package with JSON test vectors. No LLM, no API key, no server, no account.
• An FBS-owned dataset, cited. Compiled from what each chain publishes (21 CFR 101.11). Every chain carries an as-of date and a source URL, shown in the app. A validator checks every row (calories against 4·P + 4·C + 9·F within a stated tolerance). A mothership job refreshes it quarterly.
• Honest modifications. "Remove X" and "add X" are computed only where the chain publishes component-level nutrition. Everything else is marked as an estimate, or not offered.
• Published totals are the base. An item's figures are the chain's published total for the item as sold. A modification is a delta on that total from a verified component row. No item is ever composed from an assumed recipe; the only summed builds are at chains where ordering is itself build-your-own with every portion published. (Added 2026-09-19: the MVP composed Subway's protein bowls and came out 405 to 455 kcal under Subway's own totals; caught by the verification pass, data/VERIFICATION-2026-09-19.md in the repo.)
• Honest coverage. A visible list of covered chains with dates, and a plain "not covered yet" state. v1 is 25 to 30 chains and does not claim every restaurant.
• No medical claims. "Best" means ranked against the user's own diet. A disclaimer and the source citation sit on every result screen (App Review 1.4.1; a meal-planning app was rejected in 2024 for missing citations). No "blood sugar", "GLP-1" or "diabetic" language.
• Names as text. Chain names in plain text inside the app. No logos, no chain name in the app's title or icon, a non-affiliation line (guidelines 4.1 and 5.2.1).
• Privacy. No data collected. Privacy manifest in the first build.

Not in v1. Apple Health ("what is left today") is v2 and read-only; the app never logs food ([[M-056]] stands). Location-based chain picking is v2. No Android, no web product, no AI menu scanning.

Signal. For the business: 50 paid downloads in the first 60 days after launch without paid ads, and at least one piece in the macro-friendly-order format (a short video or thread built from the app's own rankings) passing 10,000 views. For the personal case: Wayne uses it at the counter at least weekly for a month. Review 2 weighs both.

Output. A ranked list of orders for the user's diet at the chain in front of them. Each has its macros, the modifications to say, an estimate flag where components are not published, the source and the date the data was checked.

Investment case

• Buyer. An adult on a named macro-driven diet (high protein, low carb, keto, Atkins, carnivore, Mediterranean, calorie cap) who eats fast food weekly and owns an iPhone. Today they use a tracker, a chain's PDF, or a subscription app built around cut / bulk / maintain. Consumer only.
• Price. $9.99 one-time, proposed; Wayne sets it. Every live competitor is a subscription ($4.99 to $9.99 a month, $19.99 to $44.99 a year, lifetime tiers $39.99 to $119.99). A one-time price under the cheapest annual plan is one of the two openings. Refresh runs on $0 runners, so one-time holds.
• Path to market (owned channels). Weak, and the main risk. The Free Beer audience is AI and business, not diet. The format that works for this content is short video and listicles ("best high-protein orders at X"); the app's rankings generate that content, one chain per piece, on X and hughmann.life, with App Store search as the base. No paid ads.
• Runs to MVP. 6 runner runs on the native chain: scaffold, dataset schema with five chains, profiles and scorer with vectors, the core flow in the simulator, the coverage screen, make build-all with screenshots and BUILD-proof.json.
• Wayne-minutes. 80 to a listed app: two reviews (13), the name and the price (4), the App Store Connect key and signing identity (10, shared with Health Horizons), about ten task reviews (50).
• Overlap with what we sell. none: nothing on freebeer.ai serves diet buyers. The nearest relative is Health Horizons, whose HealthKit package v2 of this app would reuse, and whose buyers are the same health-minded iPhone owners.

The competitive set

The assumption this started with, that nobody ranks orders by diet with modifications, is wrong.

| App | Model and price | Ratings | What it does | Where it stops | |---|---|---|---|---| | MacroMate — Fast Food Macros (indie) | Subscription; $6.99 and $44.99 tiers, lifetime $119.99 | 4.9 (37) | Pre-built macro-friendly orders at 120+ chains from official data, with modifications; offline mode; iOS and Android | Nearly this product. Goal lenses (cut, maintain, bulk, plus keto and GLP-1), not named diets. Subscription. About a year old with 37 ratings | | CalorieCap | Subscription; $4.99/mo, $24.99/yr | 4.6 (97) | Ranks matches against a calorie target; ingredient-level builder at some chains | Calorie lens only. Listed since about 2016, under 100 ratings | | MenuFit | Subscription; $9.99 to $48 tiers | 4.8 (54K) | AI: what to order at any restaurant, "healthify any meal" | Macros are AI estimates; reviewers cite 28 to 40% calorie errors against the chains' own numbers. Online only | | LeanBites | Subscription; $9.99/mo, $19.99 to $29.99/yr | 4.6 (3.8K) | Nearby restaurant meals filtered by macros and diet tags | Finds meals nearby; no ranked orders per chain, no modifications | | What2Eat | Subscription; $3.99 to $49.99 tiers | 4.7 (414) | High-protein and low-calorie picks per restaurant | No named diets, no modifications | | Cheat Day | Subscription; $4.99/mo, $19.99/yr | 4.3 (323) | 800+ menus sortable by macro; Apple Health | A sortable database, not ranked orders | | Nutritionix Track | Subscription; $5.99/mo | 4.8 (44K) | Tracker with 760+ chain menus | A logger; recommends nothing | | CalorieKing | Free with ads | 3.5 (314) | Offline lookup across 260 chains | Lookup only; last updated 2020 | | HealthyOut (dead) | Was free | n/a | Ranked dishes by diet with modifications, 2013–15 | Acquired 2015, gone; never reached scale as a business |

Every live competitor is freemium with a subscription. No paid-upfront app was found in the niche. The curated-data apps have the fewest ratings in the set (37, 97, 323); the downloads have gone to the AI products that cover any restaurant (54K, 3.8K). Demand for the *content* is plain: TikTok has discovery pages for "fast food macros", YouTube and the listicle press run the format constantly, and the chains themselves now ship high-protein menus (Chipotle, December 2025).

Where the opening is, and the case against

The opening is narrow and has three parts. 1. Named diets. Nobody offers Atkins by phase, carnivore or Mediterranean. The competitors think in cut, bulk and maintain. 2. Pay once. $9.99 against subscriptions whose cheapest year is $19.99 and whose lifetime tiers start at $39.99. 3. Numbers you can check. Against the AI leader's 28 to 40% errors: published data, a citation and a date on every screen. It is also what App Review asks for.

The case against. MacroMate already ships the core idea with four times the chains, and has 37 ratings to show for a year. Curated data has not won downloads for anyone. HealthyOut did this a decade ago and ended as an acqui-hire. FBS has no diet audience. If the app cannot clearly beat MacroMate on named diets and price, the evidence shows no business here.

The director's read. As a business alone this would be a Later. It is a Yes to the MVP for three reasons that do not depend on the market: Wayne is the user and wants it; it is the first venture on the native line, so its six runs also prove M-074's machinery for every app after it; and the cost is $0 runners plus about 80 Wayne-minutes. Review 2 is the real gate. Kill criteria to carry into Review 2: the modifications cannot be made accurate for at least 10 of the v1 chains, or Wayne is not using the simulator build's rankings himself.

What is settled and what is open

Settled (Wayne, 2026-09-19). An FBS App Store product. v1 is the 25 verified chains. Data independently verified 2026-09-25: all 881 rows compared against the chains' publications by a second pass (five chains 09-19, twenty 09-25), 0 discrepancies; reports in the repo's data/VERIFICATION-*.md. Own curated dataset, no licensed API. Apple Health is v2. Native line, kind apple-app.

Correction path. v1 ships data inside the app and corrects by app update. A signed data file fetched from freebeer.ai is a v2 option if the refresh cadence needs it; it would be the app's only network call and changes the privacy story, so it is not assumed.

Native MVP (the four steps of FACTORY § 9a). Scaffold from the M-055 pattern → dataset slice and Core scorer → one flow in the simulator (diet → chain → ranked orders) → make build-all, make core-test, make screenshots, BUILD-proof.json. Review 2 is shown screenshots and a simulator recording; TestFlight when the key exists.

The v1 chain list (research/chains-2026-09-19.md; top 30 by US units, QSR 50 2026, sources opened 2026-09-19). Of the 30, 12 publish component-level nutrition, 8 partly, 6 not, 4 could not be verified.

• First slice of five: Subway, McDonald's, Chick-fil-A, Chipotle, Five Guys.
• Modifiers on (12): those five, then Taco Bell, Domino's, Pizza Hut, KFC, Firehouse Subs, Panda Express, Wingstop.
• Limited modifiers (8): Arby's, Jersey Mike's, Dunkin', Sonic, Dairy Queen, Little Caesars, Panera, Tropical Smoothie.
• Base items only (5): Starbucks, Wendy's, Popeyes, Papa Johns, Jack in the Box.
• Only after a hand check (5): Burger King, Hardee's, Whataburger, Marco's, Jimmy John's.
• The gap that matters. Bunless, carnivore and keto ordering matters most at burger chains, and only McDonald's and Five Guys have confirmed bun and patty rows. At the other seven burger chains the app ranks base items and switches modifiers off. This bears directly on the first kill criterion: 12 chains clear the bar of 10 today, but the burger category does not.

Data rights, open. Nutrition values are facts and not copyrightable (Feist), and the rule that makes chains disclose them (21 CFR 101.11) requires neither online publication nor component data, so every component source is voluntary. The separate risk is contract: the site terms read (McDonald's, Chipotle, Chick-fil-A, Five Guys, and Syndigo for the six chains it hosts) ban copying or automated access in general words. The posture in this PRD: values are re-keyed by reading the published PDF or page into our own schema, no endpoint is called, no layout, prose or image is copied, and every chain cites its source. Settled (Wayne, 2026-09-19 ~11:00 CT): "go with the chains you have information for." v1 ships the 25 chains whose sources were opened and read, from hand-keyed published numbers with the source cited per chain. The five that could not be verified (Burger King, Hardee's, Whataburger, Marco's, Jimmy John's) are out of v1 and shown as "not covered yet"; each comes in only after its source is read by hand. The posture above (re-key, no endpoints, no copied layout or prose, citations) is the rule for every chain.

Wayne's (set 2026-09-25). Name Drive-Thru Macros; price $9.99 one-time; iPhone-only. Still his: the Apple Developer enrolment / App Store Connect key and signing identity.