The goal is unicorns, built with exceptional minds who are handed execution systems that have already been tested. AI has collapsed the cost of building, so the edge has moved to picking the right markets and executing in parallel with absolute standards. ILES is the internal engine for exactly that: a small core team, our own engines for every function of a company, and passionate domain founders spinning up multiple AI-native companies at once, then doubling down on whichever one captures its market.
Rocket Internet proved that company creation can be an industrial process. Consumer-goods companies proved that you win on the plant and the shelf, not on the formula. Alexander proved that a small, disciplined force with better systems takes empires. We take all three, for software, in a world where the product is the cheap part.
Million-dollar enterprise
Unicorn
Global company
Merged down from sixteen so one tick is one distinct advantage.
Below the floor it is an interest, not a venture.
Sector-scale outcome. The double-down candidate.
Category-defining. Everything the factory has goes here.
Six operating principles. They are not values on a wall. Each one decides something concrete about which companies get started, how many run at once, and when one is closed.
A competent operator with the right agents can take an idea to a working product in days. When that is true for everyone, the product stops being the edge. What is left is judgement: which market, which wedge, and whether we actually hold leverage there. We spend our attention on the picking and let the engine do the building.
A consumer-goods company does not win on the recipe for soap. It wins on the plant, the distribution and the shelf. We are the plant. Each company is a product on a shared line: the same spine, the same growth engine, the same economics desk, the same weekly number.
We do not run one bet at a time and hope. Several companies are spun up at once from the same engine and judged against the same standard. The portfolio picks the winner, not the pitch. Parallel is not a distraction here. It is the point.
Every internal engine is matched to the current best-in-class external standard. Research, build, design, marketing, sales, execution, organisation, economics, legal and data are each benchmarked against the best thing available outside the building. If the outside world does it better, we upgrade or replace. Nothing ships below the benchmark.
Eleven leverages, scored honestly. Below thirty percent we do not start. Around seventy percent the outcome is a unicorn. At a hundred percent it is a global company. The score is decided before the first line of code, and it is what tells us whether to spin something up at all.
Parallel bets exist to find the one that wins. When a company starts capturing its market, the core team, the capital and the engine's full attention go to it. The rest are held, merged or shut without ceremony.
Eleven leverages, no overlap. Every idea is scored against the same board before capital moves, and the score is the ceiling of the outcome: at thirty percent the company can be a million-dollar enterprise, at seventy it can be a unicorn, at a hundred it can be a global company. Below thirty we do not start.
We select on this board, and we use it on every company we run. Nothing proceeds until we are sure the leverages are actually there, not hoped for.
Enough money, or access to it, that a single bet does not have to be the whole bet.
team, mentors, network
A team that ships, mentors who have done it, and people who take the call on day one.
domain depth, non-public information
You have lived the sector and you know something about it that is not public yet.
AI, software, automation
The engines to build, market and run a company in days, at a fraction of the incumbent's cost base.
A route to the buyer that does not depend on one person's calendar.
A name the buyer already trusts, in this sector, before the first call.
Proprietary workflows, models or data that a fast follower cannot copy in a week.
processes, gates, weekly numbers
A written playbook, kill criteria set before capital moves, and a number posted every week.
A clear picture of what the sector looks like after the incumbent loses, and the will to get there.
other people's time, asynchronous work
Thesis to product in days, and agents carrying the headcount so humans carry only judgement.
A wedge the incumbent structurally cannot follow without breaking its own P&L.
Tick only what you genuinely hold today. This is the same check we run on you and on every idea, and we only proceed when the board says we should.
Not yet.
An interest, not a venture. It goes on the board and waits for a leverage to change.
We only proceed when we are sure the leverages are there. If the verdict says unicorn and you can defend every tick, message me.
This is the record so far, and the current batch. Every company here was picked with the same questions the leverage board asks: which market, which frozen incumbent, and which leverages on the board we actually hold.
Hospitality
Built to ₹1 crore ARR, then handed over.
The first company. Built to a crore of annual recurring revenue, then left to work on things with a higher ceiling. It is the proof that the operating discipline works before the AI tooling existed.
AI-run direct-to-consumer fashion jewellery.
A consumer brand where the creative, the marketing and the operations run on the engine. Product photography, campaigns and merchandising are generated and tested at a pace a conventional D2C team cannot match.
AI-native law platform for M&A and case intelligence.
Deal work and case research are judgement applied to documents, which is exactly the shape of work agents now do at cost. TRULX is the platform for M&A diligence and case intelligence, built for the firms whose incumbents bill that work by the hour.
Hyper-realistic, editorial-grade AI photoshoots for e-commerce owners.
Vogue and Victoria's Secret level output at a fraction of the cost of a shoot. Aimed at the millions of online store owners who cannot afford a studio, a model and a crew but still have to compete on the same shelf.
The IDE for learning.
A VS Code-type platform for building education apps, with personalisation built into the layer underneath. Instead of one more learning app, the environment that learning apps get built on, so each learner's model of what they know follows them across everything built on it.
We do not enter a market because it is interesting. A sector qualifies when it moves the economy, touches most people, and its leader has a structural reason to stand still. Then the leverage score says whether we execute.
Shares are US value added by industry as a percentage of GDP, 2025 Q3, from the Bureau of Economic Analysis, rounded to one decimal. Local services combines accommodation, food services and other services. Defence is national-defence spending and sits inside the government line. The board is where the next batch of companies is picked from.
Roughly 1% of value added or more. We only enter sectors that move the economy.
The sector touches most of the population. Consumer social, ad tech and speculative finance fail this test and are permanently out of scope.
The leader has a services P&L, a seat-based price, an install base or a roadmap that makes the obvious AI-native fix self-harming.
The board says at least 30%, with every tick defended by evidence, and a named domain founder who holds the sector-side ones.
The buyer exists. The product does not.
Demand is already spending money on labour, outsourcing or spreadsheets, and no software captures it. Fastest to revenue, weakest defence.
The leader owns the market and cannot use the new tool.
Seat pricing that AI destroys, a services line AI cannibalises, an install base that cannot migrate. We arrive with a fraction of their cost base and nothing to protect.
The moat is paperwork nobody wants to do.
Licences, accreditations, audit trails, procurement vehicles. Slow and expensive, and exactly the kind of barrier a factory can buy once and amortise across a portfolio.
Nobody can do it yet. Someone will.
The bottleneck is science or hardware, not go-to-market. Kept small, funded from portfolio cash flow, and required to sell something real while the hard part is still hard.
| Sector | % GDP | Weight | Gap | Who is frozen, and why |
|---|---|---|---|---|
| Real Estate & Property | 13.7% | Market gap | Assessors, appraisers and tax consultancies billing humans to read public records. | |
| Government & Public Services | 11.2% | Regulatory gap | Procurement cycles and legacy vendors with multi-year contracts and no incentive to ship. | |
| Manufacturing & Industrials | 9.5% | Incumbent gap | On-premise MES and ERP suites with six-month implementations and no path to autonomous cells. | |
| Finance & Insurance | 8% | Incumbent gap | Underwriting desks and middle offices running judgement over documents on email and Excel. | |
| Healthcare & Payers | 7.7% | Incumbent gap | Outsourcers who sell administration as billable labour, so automating it destroys their revenue line. | |
| Local Services, Hospitality & Trades | 5.3% | Market gap | Millions of small operators with no software at all, served by lead-gen marketplaces taking a toll. | |
| Construction & Housing | 4.3% | Incumbent gap | Estimating, permitting and project software built for the desk, sold per seat, priced for 2009. | |
| Defence & Critical Infrastructure | 3.7% | Regulatory gap | Clearances and accreditation as the moat, with primes optimised for cost-plus rather than speed. | |
| Freight & Logistics | 3.3% | Incumbent gap | Brokerages taking a margin on information asymmetry maintained by humans on phones. | |
| Energy, Water & Utilities | 1.5% | Regulatory gap | Regulated rate bases on fifteen-year capital cycles, and thousands of small water operators reporting compliance by hand. | |
| Agriculture & Food Systems | 0.9% | Market gap | Input dealers and co-ops as the information channel, with margin captured by whoever knows the price first. |
Engel's law, generalised. As income per person rises, the share spent on food falls and the shares spent on housing, transport, health, leisure and financial services rise. Money does not just grow, it migrates. We pick sectors inside the categories the money is migrating to, in the countries where incomes are crossing the bands.
Global figures are approximate 2024 annual spend in USD trillions, assembled from World Bank household consumption totals, the ICP consumption structure, WHO and HolonIQ, rounded to the nearest trillion. The curve shapes follow Engel's law; the points are calibrated to ICP country groups and US consumer-expenditure quintiles and are illustrative, not a dataset.
| Category | at $1k | at $3k | at $12k | at $50k | $1k → $50k |
|---|---|---|---|---|---|
| Food & beverages | $500 | $1.1k | $2.6k | $5.0k | ×10 |
| Housing, utilities & home | $120 | $480 | $2.4k | $13k | ×104 |
| Transport & mobility | $50 | $240 | $1.4k | $7.0k | ×140 |
| Health | $30 | $150 | $840 | $5.0k | ×167 |
| Financial services & insurance | $10 | $90 | $840 | $6.0k | ×600 |
| Recreation, media & leisure | $20 | $120 | $960 | $5.5k | ×275 |
| Education | $20 | $120 | $600 | $2.5k | ×125 |
Income grows fifty times across the bands. Food grows ten times. Financial services grow six hundred times. That gap is the divergence, and it is where the next sectors form.
From staples bought raw to processed, packaged, delivered and eaten out. The share falls, the money keeps rising, and it moves up the value chain.
From shelter to a place that is owned, financed, furnished, powered and serviced. Housing becomes the largest line on the household's books.
From walking and buses to a two-wheeler, a car, flights and everything financed and insured around them.
≈$10T including public spending
From emergency care paid in cash to insured, preventive, continuous care. Rises with income and again with age.
The steepest riser. Once food and shelter are covered, surplus becomes savings, credit, insurance and pensions.
≈$8T including restaurants & hotels
From nothing to a budget for travel, screens, sport and experiences. Almost entirely discretionary, so it tracks surplus income directly.
≈$6T including public spending
Rises fastest in the middle bands, where households pay privately to get their children past the state system, then plateaus where the state provides.
Pick the category the money is migrating to. Pick the subcategory inside it that is forming into a sector. Then check the leverage board. A company only starts when all three say yes.
A small core team cannot run many companies at once on effort. It runs them on tooling. The engine is ten internal tools that carry the repeatable work of picking, building, selling and running a company, so the humans can spend their attention on judgement.
5 building · 5 queued
Benchmarks are re-set every quarter against the external best-in-class.
Sector selection, incumbent teardowns, leverage scoring and the written thesis. The strategy-consulting deliverable, produced by agents before any capital moves.
Takes a thesis to production software on a shared spine: auth, billing, agent orchestration, evaluation harness. Only the domain logic is new for each company.
Product design, interface, brand identity and creative, on one shared design system. Every company ships looking like the best thing in its category, not like a prototype.
Performance ads, SEO, content, social and lifecycle email, run by agents against a measured cost-per-acquisition ceiling from week one.
Outbound, pipeline, partnerships and closing. Design partners signed before code, channels that do not depend on the founder's network.
The 100-day gates, the weekly keystone number, the kill criteria. The same checklist for every company so the process is repeatable by people who have never done it.
Roles, decision rights, the agent-to-human split, hiring, compensation and equity. Every company gets an org chart where agents are named as roles before day one.
Unit economics, capital allocation across parallel companies, cap tables and the double-down decision. Every company is priced before it is built.
Entity, contracts, IP, data protection and sector regulation. The paperwork moat, bought once and amortised across the portfolio.
Analytics, dashboards and agent evaluations across every company. The weekly number is computed, not reported, and every agent crew is scored nightly against real failure sets.
This is the whole model in five steps. It is deliberately rigid. A checklist is what makes company creation repeatable by people who have never done it before.
GDP weight, majority reach, a frozen incumbent. Then the leverage score. Below thirty percent the idea goes on the board and waits. Above it, we look for the founder.
A named sector, a named incumbent, a leverage score at or above 30%.
Someone who has lived the sector and is passionate about it. Knowledge, people, positioning and the wedge come with them. The rest of the board is filled by the factory or it is not filled at all.
A founder whose ticks on the board can each be defended with evidence.
Several companies at once on the same engine. Shared spine, shared growth engine, shared economics desk. The product exists in days; the first weeks are spent proving someone will pay.
Production live with a real customer's data flowing through it.
One keystone number per company, posted weekly. Kill criteria written before capital moved. First collected invoice inside 100 days, or the company is closed.
First collected invoice from an unrelated party.
When one company starts capturing its market, it gets the core team, the capital and the engine's full attention. The rest are held, merged or shut.
A company that is winning its market on a measured number.
Companies are started in batches, not one at a time. The engine does not care how many are running.
Every company posts the same kind of number, weekly, against a kill line written before it started.
The point of parallel is to find the company that captures its market. Then the factory goes all in.
I am neurodivergent, and I love doing multiple things at once. For a long time the advice was to pick one. This factory is my answer to that advice: a structure where running many things in parallel is the design, not the distraction.
The engine holds the standard so I do not have to hold it in my head. The domain founders hold the depth. I hold the picking: which sector, which leverage, which company gets doubled down on.
If you understand a market better than the people currently serving it, and you would rather build with a factory behind you than alone, I want to hear from you.
Founder, ILES
Neurodivergent, and built for running many threads at once. The engine exists so that the standard lives in the system, not in one person's working memory.
Not a fund. Not an incubator. A factory whose only output is companies that capture their market. We find people who are cracked enough to think they will build a unicorn, check that they hold real leverage, and give them the engines and the historical patterns that have already worked: Rocket Internet's hundred-day plan, the consumer-goods plant, and the small disciplined force that took an empire.
Alexander, for software. The point is to conquer.
No form, no deck, no application. One message: the sector you understand better than the people currently serving it, the incumbent that cannot respond, and which leverages on the board you already hold. If it is real, you get the factory.
You have lived a sector for years and you know exactly where the incumbent is stuck. Depth, not a survey of trends.
Knowledge, people, distribution, positioning, vision. You should be able to defend every tick on the board, and the board should say unicorn.
One number posted weekly against a kill line written before you started. You want that, because it is what makes the outcome real.
The sector, the incumbent, why they cannot respond, and the leverages you hold. Plain text. No deck.