How My App Is Doing (2 Month Update)

SUMMARY

Indie developer Chris shares a candid 2-month update on his AI calorie-tracking app Amy, detailing revenue growth to $1,500 monthly, AI cost reductions from $700 to $221, new features like menu scanning, and retention boosts from 3% to 10%.

STATEMENTS

Chris launched his AI calorie-tracking app Amy two months ago, initially facing high AI costs, low profit margins, and poor retention, but has since made significant improvements in sustainability.
The app now generates $1,500 in monthly recurring revenue from 148 paid subscribers at $10 per month, with the last 28 days yielding $1,968, which Chris considers strong for a young product.
AI costs dropped from $700 in the first month to $221 this month despite more users, primarily due to optimizations in model usage and caching strategies.
Perplexity Sonar, an expensive AI model costing about half a cent per call, powers initial food logging and calorie calculations for accuracy and speed, but edits were inefficiently using it until changes were made.
For simple portion edits, Chris switched to the cheaper Gemini 2.5 Flash model, costing a tenth of a cent, handling basic math and natural language processing for edge cases like "a couple bites."
Caching nutrition data in Supabase database reduced Perplexity calls by 60-70%, as users often log the same foods repeatedly, slashing costs by nearly 70% and enabling new features.
The new menu-scanning feature uses Gemini to extract and reconstruct restaurant menus from photos, allowing users to tap items for calorie logging, a unique addition Chris delayed due to prior cost concerns.
PostHog analytics tool tracks all metrics, including AI costs per model ($217 for Perplexity, $3 for Gemini), latency, retention graphs, and user behaviors like download-to-signup rates.
Week-one retention improved from 3% at launch to 8% after a month, then to 10.2% recently, serving as Chris's northstar metric to gauge app value, targeting 30-40% for success.
Key retention boosts came from UX tweaks like explicit edit buttons for photo features, adding weight tracking with progress photos, enhanced Apple Health syncing for proteins and sugars, better data export, and smarter notifications that avoid redundant reminders.

IDEAS

High initial AI costs can cripple early-stage apps, but targeted optimizations like model switching for edits and caching common queries can dramatically improve margins without sacrificing functionality.
Users exhibit predictable eating patterns, with 60-70% repeating the same foods daily, turning this behavioral insight into a cost-saving caching mechanism that exceeds expectations.
Free-form text inputs in apps create endless edge cases for programmatic handling, making AI indispensable even for simple recalculations, but cheaper models suffice for routine tasks.
Unique features like menu scanning, enabled only after cost controls, create "magic moments" that lock in users by offering irreplaceable convenience competitors hesitate to match due to expenses.
Small UX friction reductions, such as adding visible edit buttons, accumulate to meaningfully boost engagement, revealing hidden user assumptions through direct feedback.
Storing personal data like weight progress and photos increases switching costs, fostering long-term retention by making users reluctant to abandon their digital history.
Building user trust via easy data export and robust integrations like Apple Health encourages deeper investment in the app, as people feel ownership over their information.
Annoying notifications can drive users away permanently; context-aware reminders that respect logged activities prevent this, highlighting notifications as a double-edged retention lever.
Failing to capture timestamps early can hinder future features like personalized timing, underscoring the importance of forward-thinking data collection in app design.
Analytics tools like PostHog provide indispensable visibility into costs and behaviors, allowing developers to benchmark and debug issues proactively for sustainable growth.
Indie developers can achieve healthy 85% AI profit margins by iterating on costs and features, proving small teams can compete through agility rather than scale.
Retention percentages are notoriously hard to shift, but consistent small wins compound, turning a 3% starter into 10% progress toward industry benchmarks.

INSIGHTS

Optimizing AI usage by tiering models and caching exploits user habits for massive efficiency gains, transforming potential financial pitfalls into competitive advantages.
Retention hinges on reducing invisible frictions and building data moats, where even minor UX clarifications and export options cultivate user loyalty and investment.
Analytics-driven decision-making reveals hidden patterns, like repeated food logs, enabling precise interventions that scale impact far beyond initial expectations.
Unique, cost-sensitive features like menu scanning differentiate apps in crowded markets, creating habitual dependency through delightful, hard-to-replicate experiences.
Notifications must evolve from generic nudges to intelligent, context-aware prompts to avoid alienation, emphasizing adaptive personalization for sustained engagement.
Early oversights in data schema, such as missing timestamps, amplify future development costs, reinforcing proactive architecture for evolving app capabilities.

QUOTES

"For an app that's only 2 months old, I am very happy with this."
"I was not aware of how big a difference that this made. I think this was the bulk of why the bill went from $700 to $200."
"It's definitely a magic moment for me. But I think when people use this a couple times, that's when they'll go from, okay, this is a cool feature to I can't use an app that doesn't have this."
"It's the best signal that I have found to answer the question, have I built something truly valuable?"
"The less friction that an app has, the more likely people are to stick with it."

HABITS

Chris focuses on one productivity app per video, dedicating content to deep dives into its development and updates.
He regularly talks to users to identify pain points and frictions, using feedback to prioritize UX improvements.
Chris tracks metrics obsessively with tools like PostHog, monitoring costs, retention, and behaviors daily to inform decisions.
He iterates on features based on cost feasibility, delaying ambitious additions like menu scanning until margins allow.
Chris builds in public, sharing transparent updates on revenue, challenges, and learnings via videos and social media daily.

FACTS

Amy's AI costs fell 68% from $700 to $221 monthly, despite user growth, through model optimization and caching hitting 60-70% of queries.
Week-one retention rose from 3% at launch to 10.2% after targeted changes, with successful apps aiming for 30-40%.
Perplexity Sonar charges about $0.005 per call for web searches, while Gemini 2.5 Flash costs $0.0005 for simple edits.
148 subscribers at $10/month drive $1,500 MRR, with 28-day revenue at $1,968 for the 2-month-old app.
Overall profit margins improved to 85% on AI operations, factoring in hosting, after initial 50% struggles.

REFERENCES

PostHog: Analytics tool for tracking costs, retention, and LLM metrics.
Supabase: Database used for caching nutrition data to avoid repeated AI calls.
Perplexity Sonar and Gemini 2.5 Flash: AI models for calorie calculations and edits.

HOW TO APPLY

Assess AI costs by logging per-model expenses with analytics tools, identifying high-frequency calls like edits that can switch to cheaper alternatives.
Implement caching for repeated inputs, such as common foods, using a database to store results and reduce expensive API hits by querying user patterns first.
Gather user feedback on frictions, then add explicit UI elements like edit buttons to clarify features and lower abandonment rates.
Integrate data storage features like weight tracking and progress photos to increase user switching costs and encourage habitual use.
Refine notifications to check logged activities before sending reminders, preventing annoyance and maintaining engagement levers effectively.

ONE-SENTENCE TAKEAWAY

Iterative cost optimizations and user-centric tweaks transformed Chris's struggling calorie app into a sustainable, engaging product.

RECOMMENDATIONS

Prioritize caching and model tiering early to manage AI expenses, ensuring scalability before adding resource-intensive features.
Use week-one retention as a core metric to validate product value, targeting incremental gains through UX polishing.
Incorporate user data exports and health integrations to build trust, prompting deeper engagement without fear of lock-in.
Develop context-smart notifications based on logged behaviors and timestamps to boost retention without overwhelming users.
Leverage analytics for real-time benchmarking of costs and metrics, enabling proactive fixes for emerging issues.

MEMO

Two months after launching Amy, his AI-powered calorie-tracking app styled like Apple Notes, indie developer Chris RaRoque is breathing easier. What began as a frictionless tool for typing meals and auto-calculating calories has evolved amid early hurdles: sky-high AI bills and users vanishing after a week. Now, with monthly recurring revenue hitting $1,500 from 148 subscribers, Chris sees glimmers of viability in a market dominated by bloated fitness trackers.

The pivot came through ruthless cost-cutting on the app's AI backbone. Initially, every food log or edit triggered Perplexity Sonar, a premium model that scoured the web for nutrition data at half a cent per call. Chris slashed that by rerouting simple portion tweaks—like halving a burrito—to the budget-friendly Gemini 2.5 Flash, which handles math and edge cases for a tenth the price. But the game-changer was caching: storing past queries in Supabase so repeat entries, which make up 60-70% of logs given users' routine diets, bypass expensive searches altogether. The bill plummeted from $700 to $221, unlocking margins up to 85% and freeing bandwidth for innovation.

Emboldened, Chris rolled out menu scanning, a feature he eyed warily before. Snap a restaurant menu, and Gemini extracts items, descriptions, and sections into tappable entries for instant logging—a "magic moment" absent in rivals wary of added costs. He credits tools like PostHog for this clarity, its dashboards breaking down latencies, model spends ($217 on Perplexity, mere $3 on Gemini), and retention curves. Week-one stickiness climbed from a dismal 3% to 10.2%, inching toward the 30-40% benchmark of hits like MyFitnessPal.

Behind these gains lie deliberate retention plays. Chris chatted with users, uncovering oversights like the obscured photo-edit option; a prominent button fixed that, part of 10 micro-UX tweaks easing daily use. He added weight tracking with progress photos, creating a data journal that hikes switching costs. Smarter exports to Apple Health now include proteins and sugars, affirming user data ownership. Notifications, once intrusive, now skip reminders for already-logged meals, with future plans for pattern-based timing after fixing timestamp gaps.

Not a unicorn yet, Amy hums with sustainability—revenue rising, costs tamed, a roadmap clear. For Chris, who builds productivity apps in public, the thrill lies in weekly learnings: proof that solo developers can thrive by obsessing over frictions, feedback, and fundamentals. As he eyes smarter features, Amy embodies indie grit in an AI-saturated world.