Table of Contents

minimi 2.0 Investment Report

minimi 2.0 Investment Report

Category: A Mac app that passively records what the user reads, hears and types, stores it as memory on the device, serves it to Claude and other AI tools through MCP, and tracks unfinished commitments (“open loops”)

Company Stage: Pre-seed or seed. Institutional backing is reported, but no round has been publicly announced

Founders: Jay Gadekar (Co-founder and CEO) and Ojasvika Sahu (Co-founder and Chief Design Officer)

Headquarters: Bengaluru (Indiranagar). LinkedIn lists San Francisco (see Founder section)

Funding: Boundless Ventures is an investor, per the founder. The amount and terms are not publicly disclosed

Business Model: Freemium consumer and prosumer subscription. The paid tier’s price is not publicly disclosed

Product Hunt Launch Date: September 24, 2026 (minimi 2.0). Earlier launches: Shram (July 9, 2024), Shram 2.0 (February 20, 2025), minimi (June 5, 2026)

Report Date: September 29, 2026

Investment Metric	Assessment
Venture Potential	46/100
Unicorn Path	Conditional
Valuation Attractiveness	Not Assessable
Evidence Confidence	34/100
Final Decision	Watch

Executive Summary

minimi runs in the background on a Mac. It reads screen content through Apple’s accessibility framework, without taking screenshots, and captures call audio. It stores this as memory on the device and offers it to Claude, ChatGPT, Gemini and agent tools through an MCP server (website). Version 2.0 adds an “open loops inbox” that finds commitments in Gmail, Slack, WhatsApp and Google Meet and tries to close them (Product Hunt forum). Each capability is presented as an “AI cat”: Cotton handles memory and Melody handles commitments.

The target user is a heavy AI user on a Mac who switches context often. That is mainly founders, operators and knowledge workers who already live inside Claude.

The product addresses a real gap: AI models only know what they’re told or what they’re connected to. The founders argue that capturing everything in the background covers the rest, including WhatsApp, LinkedIn, Figma and tabs, without setting up integrations (podcast transcript). Keeping memory on the device is a genuine privacy advantage.

The strongest positive signal is early organic pull. The founders say minimi began as a side project, reached about 1,000 users in more than 20 countries within weeks, and started converting users to paid plans. The CEO says users “more than doubled in the past 4 weeks,” with 95% coming from the co-founder’s social content (LinkedIn). All of this is company-reported.

The biggest concern is the history of this category. Rewind, the most prominent Mac product that captured everything on screen, was acquired by Meta and then shut down (Limitless). Apple, Anthropic and OpenAI are all building memory into their own products. minimi is the company’s third product direction in about two years, and nothing about its revenue is public.

Decision: Watch. The product idea is interesting and there are signs of organic pull. Several gaps prevent formal due diligence: pricing is undisclosed, revenue is unknown, the product only works on Mac, a single co-founder drives most user acquisition, and the company keeps pivoting.

Product Overview

The product addresses two related problems. The first is re-explaining your situation to an AI every session. The second is letting commitments slip across a scattered mix of communication tools.

minimi captures activity through macOS accessibility APIs and system audio. The founders say this is faster, cheaper and more accurate than taking screenshots (podcast). It breaks that content into small pieces (embeddings) and stores them in a vector database on the user’s Mac.

The website says “none of your data is stored on our servers” (website). The CEO, however, has said that memory-building makes one API call to a paid Gemini server, and that fully local processing is a future goal. Data may therefore pass through a third-party model even though it is not stored. This should be clarified before the “no cloud” claim is accepted.

Pricing is only partly known. The founder described a free tier of about five MCP calls a week, with a higher paid tier whose price he did not state. He said the limits are still being tested to find the point that drives conversions. A community offer gives one month free (Product Hunt). No pricing page was found.

Platform: macOS only. The founders describe Windows support as “a black box” they have not yet looked into.

What it replaces: meeting-notes apps such as Granola (the founders pitch it this way), manual second-brain tools such as Obsidian, AI memory features, and individual integrations for each tool.

Founder and Team Assessment

Jay Gadekar studied architecture at Sir J.J. College of Architecture (LinkedIn notes “All India Rank 10”). He worked as an architect at Udichi from 2018 to 2023 and spent a year as a self-employed “tinkerer” in 2024–25 (LinkedIn).

Ojasvika Sahu is also a trained architect. She leads design and, by the CEO’s account, almost all user acquisition.

Both founders describe themselves as non-technical and self-taught (Razorpay Rize profile). That profile also reports that they:

Bootstrapped for about three years.
Launched an MVP in 2023 that “didn’t land.”
Fired their CTO right before signing a term sheet.
Pivoted from Shram.io to Shram.ai in 2025.
Once declined an investment offer.

They are resilient and show strong product taste. They have no prior exits, and their technical depth depends on a small hired team.

Team: The CEO described “a tiny but competent team of 4” in early 2026. Current job postings include a frontend engineer (₹40 LPA) and a founder’s-office generalist (₹45 LPA) (LinkedIn post).

Headquarters conflict: LinkedIn gives San Francisco for both the company and the CEO. Job postings, events and the podcast all place the team in Bengaluru. This report treats Bengaluru as the operating base. The legal entity was not verified. The CEO has also described earlier delays with compliance filings that postponed fundraising.

Key-person risk: high. The CEO attributes 95% of users to one co-founder’s content.

The founders have also said they see the company as “a stepping stone” toward a future venture in modular city-building (Rize). That raises a question about long-term commitment to this category.

Founder Assessment: The founders are resilient, design-led and strong at content-driven distribution, but technical depth, focus and long-term commitment to this category are unproven.

Market Opportunity

The narrow initial segment is Mac-based knowledge workers who pay for Claude or ChatGPT and switch between many communication channels. Founders, consultants, product managers and investors are the main examples.

A bottom-up estimate, using analyst assumptions:

Assume 2–5 million heavy paid AI users on Macs worldwide.
Assume 5–10% would install an app that records their screen and audio continuously.
Assume $150–250 a year per user, similar to Granola-style prosumer prices.
That gives about $15M–$125M a year.

That range is enough for a solid niche business. It is not enough for venture scale unless minimi expands.

Possible expansions:

A team “company brain” made of each member’s local memory, which the co-founder has demonstrated (podcast).
Windows support.
Licensing its retrieval technology.

Timing works in both directions. Demand for AI context is high, but Apple, Anthropic and OpenAI are all investing in memory, and the CEO acknowledges this.

Traction and Growth Signals

Launch attention: All four launches ranked in the top four of the day. Shram, Shram 2.0 and minimi each ranked #2; minimi 2.0 ranked #4 on September 24, 2026 (Product Hunt). The minimi 2.0 forum post has about 50 upvotes. These results show the team is skilled at launching. They do not show product-market fit.

Company-reported signals:

About 1,000 users in more than 20 countries as of about July 2026.
“Paying users” from several countries after eight weeks.
More than doubling over a four-week period.
A launch article with about 50,000 readers.
Retrieval results about 50% better than state of the art on the BEAM benchmark.

These come from the podcast and the CEO’s LinkedIn. None is verified, and the benchmark has not been independently reproduced.

Missing: paying user count, conversion rate, revenue, retention, daily active usage, and the paid tier price.

Funding signals: Boundless Ventures is named as an investor by the founder on the podcast, and Tracxn lists Shram in its portfolio (secondary source). A contact-data site shows both “$9.1M total funding” and “never raised” (Prospeo). That source contradicts itself and is disregarded.

Traction Assessment: There are real early signs of organic pull, but the user base is small and commercial results are unknown.

Competitive Position

Direct competitors: Screenpipe (source-available, cross-platform), and “personal context” apps built on memory APIs such as Mem0.

Recent exit in the category: Rewind/Limitless, acquired by Meta and then shut down (Limitless).

Indirect competitors: Granola and other meeting-notes tools, Obsidian and other second-brain tools, and the built-in memory in Claude and ChatGPT.

Platform threats: Apple Intelligence’s on-device personal context, and Anthropic or OpenAI desktop apps that can read the screen.

Differentiation: capture through accessibility APIs without screenshots, on-device storage, independence from any single AI model through MCP, and a warm, character-based design.

Switching costs grow as memory accumulates. There is no network effect yet, although the team-brain idea could create one.

Six-month test: If Apple or Anthropic shipped ambient on-device memory for Mac, minimi’s remaining arguments would be independence across models and retrieval quality. The first matters to a subset of users. The second is not independently verified. Its dependence on macOS accessibility permissions is a structural platform risk.

Defensibility Assessment: Low.

Business Model and Economics

minimi uses a freemium subscription that limits MCP calls. The paid price is undisclosed. If it is in the $10–25 a month range typical for the category, which is an assumption, a subscriber is worth about $120–300 a year before payment and app-store fees.

Costs should be low because storage and capture happen on the user’s machine. The main variable cost is the Gemini call used during memory-building, and the company aims to remove it. That structure could support high gross margins, but actual margin is not disclosed.

Acquisition today comes almost entirely from founder content, which costs little but does not scale easily and depends on one person. Retention for passive-capture apps is unproven. Users may uninstall over privacy worries, performance impact or low perceived value.

Revenue per user is capped at consumer levels unless a team product is built.

Unicorn Path

The assumed multiple is 8x ARR, typical for a prosumer subscription business with high gross margin but higher churn than B2B software.

Required ARR = $1B ÷ 8 = $125M

At an assumed $180 a year net of fees, that requires about 700,000 paying subscribers, which is several hundred times the current reported user base. With a 5% free-to-paid conversion rate, it implies roughly 14 million active free users. For a Mac-only app that asks for broad permissions, that is not a realistic target.

A team or enterprise “company brain” changes the numbers:

At $50K a year per customer: about 2,500 enterprise customers.
At $15 per seat per month: about 700,000 seats spread across companies.

Reaching either would require:

Security and compliance work for enterprise buyers.
Admin controls and filtering of personal data.
Windows support.
Sales capacity.
Proof that retention holds up in team settings.

Unicorn Path: Conditional. It depends on moving from a Mac consumer utility to a team knowledge platform.

Valuation Assessment

Known funding: Boundless Ventures is an investor per the founder, and Tracxn lists Shram in its portfolio. The amount, instrument, valuation and date are all undisclosed. A reference to “fundraising ops” in a job posting suggests the company is currently raising or plans to.

Comparables:

Meta’s acquisition of Limitless, which confirms strategic interest in the category; terms not disclosed.
Screenpipe, the closest private peer; no terms available.

Valuation Attractiveness: Not Assessable. Pricing it would require:

Paid tier pricing
Paying subscriber count and MRR
Conversion and cohort retention
Inference cost per user
Burn and runway
Prior round terms and cap table
Terms of any current round
The legal entity and where it is incorporated

Key Risks

Platform bundling. Apple, Anthropic and OpenAI can offer native ambient or desktop memory and remove the need for minimi.
Unproven willingness to pay. Pricing, paying user numbers and revenue are unknown, and the founder says most non-paying users are satisfied with the free limits.
Repeated pivots. Shram.io became Shram.ai, which became minimi. A second product (Shram) is still maintained. Focus is unclear.
Distribution depends on one person. About 95% of users come from one co-founder’s content.
Privacy and security. Continuous capture of every screen, call and message is sensitive, and the “no cloud” claim conflicts with the disclosed Gemini API call.
macOS permission dependency. Changes to Apple’s accessibility or privacy rules could break the product.
Mac-only reach. This limits the consumer market and makes enterprise sales harder.
Unverified technical claims. The benchmark results are self-reported.
Founder commitment. The stated long-term interest in a different venture raises questions about staying in this category.

Final Assessment

Venture Potential: 46/100

Category	Score
Market Size and Expansion Potential	11/20
Traction and Growth Evidence	6/20
Founder and Team	8/15
Product Strength	7/10
Distribution Potential	7/15
Business Model and Economics	4/10
Defensibility	3/10
Total	46/100

The strongest parts of the case are product design, the privacy-first architecture, and organic content-led pull. The weakest are defensibility against platform owners, unknown monetization, and a market limited to Mac users.

Evidence Confidence: 34/100

Verified: founder identities and backgrounds, the product’s existence and features, Product Hunt launch history, and hiring posts.

Company-reported: user numbers, growth, paying users, the benchmark results, and Boundless as an investor (partly supported by Tracxn).

Conflicting: headquarters (San Francisco vs Bengaluru), the data-flow claims, and unreliable funding figures.

Unavailable: price, revenue, retention, margin, burn, the round amount, and the legal entity.

Final Decision: Watch

The product is thoughtfully designed and shows early organic interest. It is not a Pass, because a team memory layer that users own could become meaningful and the team ships quickly. It is not DD, because there is no evidence yet of monetization, retention, or defensibility against platform owners.

Upgrade Conditions

These would move the decision to DD:

Published pricing and verified MRR of at least $50K–100K.
At least 5,000 paying subscribers, or 90-day paid retention above 60%.
A team or company-brain product with paying organizations.
Acquisition channels beyond one founder’s content.
Independent verification of retrieval benchmarks.
Windows support or a clear plan to expand beyond Mac.
Transparent data flows and a published security review.

Downgrade Conditions

Apple, Anthropic or OpenAI ship equivalent native ambient memory.
Another product pivot.
User growth stalls after the launch.
A privacy incident, or evidence that data leaves the device beyond what is disclosed.
Founder attention moves to other ventures.

Questions for Further Diligence

What is the paid tier price, how many paying subscribers are there, and what is current MRR?
What are 30-, 90- and 180-day retention rates for free and paid users, and how often is the app uninstalled?
Exactly what data is sent to Gemini or other third-party servers, and is any of it retained?
How are the BEAM and LongMemEval results measured, and can a third party reproduce them?
What happens to capture if Apple restricts accessibility-API access or tightens permissions?
How many active users came from sources other than the co-founder’s content?
Is Shram still maintained as a separate product, and how is the team’s time split between Shram and minimi?
How much was raised from Boundless and others, on what instrument and cap, and when?
What legal entity holds the IP, and is it in India or the US?
What is the plan for a team or enterprise product, including personal-data filtering and admin controls?
How do the founders view the long-term commitment to minimi compared with the stated Modularity ambition?