पाठशाला Pathshala · मन Man, The founder · Lesson 03 · Start

Quitting your job: timing, savings and the family conversation

Resigning is not the brave part. The brave part is deciding the trigger in advance, computing the household’s runway honestly and telling the people whose lives it changes before it changes them.

Pathshala, The Founder Library · 11 October 2026 · 9 min read

The resignation is the part that gets told at dinner parties. The parts that decide whether the story ends well are quieter: what had to be true before the letter was written, how many months the household could carry before anyone had to find out, and whether the people carrying it were asked or informed. This lesson is about those three.

It is written with some care, because the subject is not the founder alone. A job supports other people and a company, for its first year or two, mostly does not. The arithmetic below is simple. The conversation at the end is not, and no amount of conviction about the idea makes it easier for the person on the other side of the table.

The trigger: what has to be true, not what day it is

Founders who pick a date resign on the date whether or not anything has been learned. Founders who pick a trigger resign when a specific thing is true, and the thing is almost always evidence. Twenty customer conversations held and a count of how many have already paid to solve the problem. A first order with money attached. A co-founder who has committed in writing. A number of users who came back on their own. The lesson on [the first ninety days](/library/founders-first-ninety-days) sets out the quarter in which this evidence is produced, and the first six weeks of that quarter can be done in the evenings and on weekends while the salary continues.

So the trigger worth writing down is this: leave when the work the company needs can no longer be done in the hours a job leaves. For most founders that moment arrives somewhere between the problem being confirmed and the first version being built, because interviews fit around a job and delivering a product by hand to the first users does not. Before that moment, quitting buys nothing but risk. After it, staying costs the company the one thing that cannot be bought back, which is weeks.

Paul Graham’s list of reasons not to start a startup ends with the observation that a job is the default and that defaults are powerful precisely because they operate without conscious choice. The trigger is how a founder replaces the default with a decision. It also protects against the opposite failure, which is the founder who resigns on conviction alone in week one and spends months of household money finding out what the evenings would have found out for free.

Personal runway: the household’s arithmetic

Company runway is covered in [its own lesson](/library/runway-how-many-months-you-really-have). Personal runway is the same idea applied to the household, and it is the number most founders have never computed because the bank balance looks like enough. It is not a balance. It is a balance divided by a monthly cost, and the result is a month on the calendar.

Start with the savings you would actually draw on. Liquid only: bank balances, fixed deposits that can be broken, liquid funds. Not the flat, not the PF, not shares you would be ashamed to sell at today’s price. Subtract a reserve you will not touch for the company under any circumstances, because a medical or family emergency does not wait for product-market fit. Then the household’s cost a month, taken from twelve months of statements rather than memory: rent or EMI, food, school fees, parents’ support, insurance premiums, transport, the subscriptions nobody remembers. Subtract a partner’s take-home and anything the company will pay you, which in year one is usually nothing. What is left is the monthly drain, and savings after the reserve divided by the drain is the number of months. Eighteen is the figure founders who have done it twice tend to name, and the reasoning is that twelve months is roughly how long it takes to find out whether a company works and six more is what it takes to find a job if it does not.

The last two sliders turn the calculator from a verdict into a plan. If the months of cover are fewer than the months you want, the gap is a sum of money, and that sum divided by what you save each month while employed is the number of months longer to stay. That gives a quit date computed rather than felt, and a date computed this way is far easier to put in front of a family than one chosen because a friend’s company just raised.

What is not runway: PF, gratuity, options and the group policy

The provident fund is retirement money. Treat it as such. The rules on early withdrawal have been changing: at the 238th meeting of its board on 13 October 2025 EPFO decided to lengthen the window for premature final settlement from two months of unemployment to twelve, to lengthen the window for final pension withdrawal from two months to thirty-six, and to earmark a quarter of a member’s contributions as a minimum balance to be kept in the account. The direction is clear even as the detail settles: the state would prefer that money stay where it is, and so should you. Checked October 2026; confirm the current rules on the EPFO site before relying on any of them.

Gratuity has a five-year line. Under Section 4 of the Payment of Gratuity Act gratuity is payable on resignation only after not less than five years of continuous service with the employer, with exceptions for death and disablement. A founder resigning at four years and eight months leaves roughly half a month’s salary per year of service on the table. Check the date before choosing the date.

Unvested options vanish. Vested ones have a clock and a tax bill. Whatever has not vested on the last day is gone, and vested options usually have to be exercised within a window set by the plan, with a tax cost at exercise. Read the plan document and ask the company’s finance team what the window and the tax are before the resignation is handed in, not after.

The group health policy ends on the last day. India still pays for a large share of its healthcare out of pocket: the National Health Accounts put out-of-pocket spending at 39.4 per cent of total health expenditure in 2021-22, down from 62.6 per cent in 2014-15 but still the kind of share that turns one hospital admission into a year of runway. IRDAI’s guidelines on migration and portability give a member leaving a group policy the option to migrate to an individual or family floater policy, subject to underwriting, with credit for waiting periods already served. Use it, or buy a separate family floater while still employed and healthy, and if anyone depends on your income buy term cover at the same time. Premiums belong in the household cost line above.

A bank balance is not a runway. A balance divided by what the household costs each month is a date, and the date is what the family needs to hear.

The family conversation

Graham is unusually plain in the ninth of his sixteen reasons: he would not advise anyone with a family to start a startup, not because it is a bad idea but because the risk is carried by people who did not choose it. In India that observation weighs more. Households are more often joint in their finances, parents are more often dependants or co-signatories, and a single salary frequently supports more than the people living under one roof. The people the decision affects therefore include a spouse or partner, children old enough to notice, and parents, and sometimes a sibling whose own plans rest on the same income.

Three rules, each learned by founders who broke it. Ask, do not announce. A conversation that opens with a decision already made is not a conversation, and the person hearing it knows. Open with the numbers instead: what the household spends, what it has, how many months that is, and what the work so far has shown. Bring the end date, not the upside. Nobody who loves you needs to hear about the exit. They need to hear what happens if it does not work: the month by which, if a specific number has not been reached, you will go back to a job, and the salary you expect to find. A founder who can name that date is asking for a loan of time with a term. A founder who cannot is asking for faith. Say what changes for them. Income, weekends, holidays, the hours you are present, the help you can give with parents or children, the plans that are postponed. Then say what does not change: the school, the EMI, the reserve that is never touched. People can carry a great deal when they know its shape.

Give the partner a real veto and mean it. A company started over someone’s objection is a company with a second front at home, and it will lose on at least one of them. If the answer is not yet, the honest response is to ask what would have to be true, because the answer is usually a number or a date and both can be worked toward. If the answer is no, the alternatives Graham suggests, a consulting practice that grows into a product or a role in someone else’s early company, are not consolation prizes. They are how a large share of good founders actually started.

Agree one more thing before the conversation ends: a monthly check-in, on the same day as the runway calculation, in which the founder says in two sentences where the money is and where the number is. Not a presentation. A sentence each. The check-in is what turns a single frightening conversation into a shared project with information in it, and it is the single thing founders who have been through this say they would not do without.

Leaving well

Give the full notice and work it. Hand over properly, write the document nobody asked for, and leave a manager who would rehire you, because that manager is the job you may need in month eighteen and may also be the first customer, the first reference or the first angel. Do not take anything that is not yours: code, customer lists, documents, people. Read the contract for intellectual property and non-compete clauses before the last day, and if the company you are starting is anywhere near the employer’s business, get an hour of a lawyer’s time and get the position in writing. If a dispute with a former employer ever comes, the paperwork from this week is what decides it.

Then take two weeks off before day one. Not to think about the company; to sleep, see the people above and arrive at the ninety days rested. Founders who go straight from the last day of one thing to the first day of the next carry the fatigue of the job into the company and call it drive. It is not drive. It is a debt, and it comes due in month four.

A monthly ritual, in fifteen minutes

On the first working day of each month, before the company’s own numbers, update the household’s: savings, cost, partner’s income, what the company paid you. Read the two dates off the figure above. Compare the second to the end date you named in the conversation. Then say the state in two sentences to the person who agreed to it. If the date has moved closer, say so. If a stopping condition has been met, the conversation that follows is the one you promised to have, and having it on time is the whole of what honour means in this situation.


Nothing here is legal, tax or investment advice. The rules on provident fund, gratuity, options and insurance change; the figures above were checked in October 2026 and the sources are below. Read the family section twice.

Sources

  1. Paul Graham, Why to Not Not Start a Startup, March 2007
  2. Press Information Bureau, Dr Mansukh Mandaviya chairs 238th meeting of the Central Board of Trustees, EPF, 13 October 2025 — Final settlement window 2 to 12 months; pension withdrawal window 2 to 36 months; 25 per cent minimum balance.
  3. Payment of Gratuity Act, 1972, Section 4, as reproduced by the Income Tax Department
  4. Ministry of Health and Family Welfare via PIB, National Health Accounts estimates 2020-21 and 2021-22, October 2024 — Out-of-pocket expenditure 39.4 per cent of total health expenditure in 2021-22, from 62.6 per cent in 2014-15.
  5. IRDAI, Guidelines on Migration and Portability of Health Insurance Policies, IRDAI/HLT/REG/CIR/003/01/2020, January 2020