Squaring the circle on employee equity

2026-05-01T00:00:00+00:00

People work for startups for many reasons. They may like being part of a smaller team, getting more responsibilities, working at a faster pace with less red tape, feeling more connected to the mission. But despite regular advice to ignore equity, many employees greatly value access to startup equity and dream of an outsized outcome. This is why every startup talks up their equity package during their hiring process.

Having worked at startups both as a founder and an employee, I have seen employee equity through a few different angles. I've received options and RSUs, seen two companies exit, walked away without exercising, crafted an equity compensation plan, and calculated payouts in an acquisition.

My conclusion: the current tools at our disposal to give employees equity exposure suck, and most employees do not understand the limits of them. In this post, I share the issue with a few different structures and weigh the best options available.

DISCLAIMER I am not a lawyer, and I certainly am not your lawyer. This post simplifies many concepts, and offers a layman's understanding of a very complicated topic. Please consult with a tax professional for any questions about your taxes and startup equity. As Microsoft lawyers are fond of saying, this post is for entertainment purposes only</a>. </blockquote>
What's wrong with employee equity?</h2>
A three-month exercise window is unworkable</h3>
The standard instrument for giving employees equity in the United States is the Incentive Stock Option ("ISO"). An ISO is an option, which means that it gives the recipient the right to buy the stock at a given price. Typically, ISOs are priced at Fair Market Value (FMV) so they have no economic benefit as far as the IRS is concerned when they are received. This is a very helpful feature: startup equity is illiquid, so giving employees straight up stock could potentially lead to a large tax bill with no liquidity to cover it.
The advantage of ISOs is that they defer taxation until the gain is realized (i.e. the shares are sold) rather than when the option is exercised, and allow employees to treat the full proceeds as capital gains. On paper, this makes it an ideal instrument for employee equity: a tax bill isn't due until a liquidity event.

Note on AMT. When ISOs are exercised, the delta between the strike price and FMV is added to income for Alternative Minimum Tax purposes, even though no stock has been sold. For employees exercising at a meaningful FMV, this can mean a real tax bill on phantom gains, with no liquidity to pay it. </blockquote>
But ISOs have a serious flaw: once someone ceases to be an employee (either voluntarily or not), they typically have three months to decide whether to exercise or surrender the ISOs. This exercise requires committing funds the employee may not have to purchase equity in a speculative enterprise sometimes years before the shares become liquid (if they ever do).
With companies staying private longer and longer, employees are more likely to need to leave before the company gets acquired or IPOs. Their life circumstances may change, they may move, or they simply may need a change. Whatever the reason, leaving presents them with an impossible choice: lock up meaningful amounts of money in company equity, or surrender the options you worked hard to earn.
I faced this choice when I left Pebble, a company I joined as a very early employee. Exercising my options would have cost me substantially all of my savings. I was pessimistic about the prospects of the company, so I surrendered my options. It felt terrible, but it was the right decision as my options turned out worthless in the Fitbit acquisition. Not all of my colleagues were as lucky.
Zach Holman, an early employee at Github, wrote colorfully about it in his post, Fuck your 90 day exercise window</a>.
A ten year exercise window ... for a price</h3>
Zach's article preaches a 10-year exercise window. Over the past decade, his and other people's writings have increased awareness of the issue and extended exercise windows have become more popular. Companies often present them to employees as getting 10 years to exercise their options after leaving the company. This sounds great! Unfortunately, it comes with an important caveat.
Here is how the extended exercise window works: when an employee leaves the company, their ISO still has a three-month exercise window. When that expires, they automatically convert to Nonqualified Stock Options ("NSO"). NSOs can have a much longer exercise window, which is how the 10 years window is achieved.
But NSOs are much worse than ISOs. When they are exercised, the delta between the strike price and the fair market value at the time of exercise is treated as ordinary income. This creates difficulties for both the employee and for the company.
For the employee, this makes exercise even more difficult. They not only need to put up funds for the exercise, they face a potentially large tax bill. But waiting for a liquidity event to exercise means the full proceeds will be taxed as ordinary income. Companies rarely explain all these details to employees, so they rarely understand the mechanics behind the exercise window.
For the company, NSOs are burdensome as well. When the NSO is exercised or liquidated, the company needs to report the ordinary income on form W-2 no matter how long ago a person may have left the company. They also need to make all applicable withholdings. This is costly (payroll tax, social security tax, medicare tax, ...), error prone (how do you know what to withhold if you do not know what the employee has paid already?), and administratively difficult. Imagine needing to process payroll for ex-employees who have since moved to a country where the company does not have a legal presence!
When Memfault was acquired, paying out all NSOs turned out to be very complicated. I cannot imagine the problems a larger company encounters.
Some companies now create guardrails to limit the number of NSO holders. For example, they may extend the exercise window by 1 year for every 2 years of service rather than automatically set it to ten years. Or they may only offer it to key employees.
RSUs only make sense at later stages</h3>
Later stage companies typically move to offering double-trigger RSUs. They vest on a similar schedule, but do not settle until some liquidity event (the second trigger). Similarly to options, this defers taxation to the liquidity events rather than on vesting.
They also solve other problems that later stage companies run into. RSUs solve other problems down the line: ISOs have a $100k limit, and AMT on exercise. When a company is so valuable that exercising options becomes practically impossible, RSUs are a good option.
Unfortunately, they run into the same problem as NSOs: they are taxed as ordinary income when the second trigger hits, and they trigger a payroll + associated tax for the company.
They also offer less opportunities for optimization to savvy employees. Exercising an option early lets you start the clock on long term capital gains. RSUs give you no such option. For early stage companies, they are strictly worse than options.
Options and RSUs expire</h3>
At this point, you may think the solution is simply for employees to keep the ISOs and stay at the company for as long as it takes to see a liquidity event. But even this approach has pitfalls.
Options and RSUs by law have an expiry date, typically no more than 10 years from issue. As companies stay private longer, employees bump into this limit more and more often.
In the best case, companies come up with a huge amount of money and do a tender offer. This allows employees to sell enough stock to cover exercise costs and tax bills. Stripe, SpaceX, Databricks, and many others have done tender offers in recent years.
But in the worst case, they simply let the options expire. This has happened! For example, Foursquare let the stock grant for over 100 former employees expire in 2023</a>. All of a sudden, the equity they thought might have been worth a life changing amount has turned to dust.
Loans are dangerous</h3>
Another solution people sometimes turn to is loans backed by vested equity. When the employee exercises their ISOs, the company loans them the money for the exercise. No money changes hands (on paper it goes from the company to the employee back to the company), and the employee gets the stock outright. In the happy case, it seems like a great deal!
If anything goes wrong however, loans can become very dangerous. Consider this scenario:

You were a trusted early employee at Startup Inc. You believed in the company's prospects and its mission, so you decided to exercise all of your equity. The company loaned you the full amount of the exercise so you did not have to put money down. Things did not work out as planned, and now the company is getting taken over by creditors. </blockquote>
What happens next? None of the scenarios are good:

The board members you knew and trusted all resigned, and were replaced by people from the creditors. They want their money back! They'll go after you for the full value of the loan.</li>
The company forgives the loan... but in some cases it is a taxable event! Now you owe the IRS some money when in fact you received nothing.</li> </ol>
Famously, some Bolt employees found themselves in a tough spot</a> after the companies fortunes turned and they were laid off. This is why Fenwick says taking out such loans is "playing with fire"</a>. Don't do it!
Exotic setups</h3>
Some companies have come up with creative setups, put in place to solve problems specific to them.
For example: Automattic, makers of wordpress, created a class of stock called A12 stock that employees can buy and sell in regular windows. A third party estimates their value on a regular basis, and the company repurchases shares from those who want them at that price.
It's an elegant solution for a company which has been private for over two decades, with no liquidity event in sight. But for a traditional venture-backed company it offers no real advantages.
You can read more about A12 stock on Automattic's website</a>.
What companies can do</h2>
Unfortunately, there are no good solutions to the issues above. The best companies and founders can do is be thoughtful about the trade-offs at hand, and carefully craft an equity plan that reflects their goals and values.
Too often, entrepreneurs just take a form equity plan provided by their law firm. These forms are typically employee unfriendly, it's the lawyer's job to protect the company!
If I were to do it again, here are some things I would consider.
Issue RSAs until the first priced round</h3>
At the earliest stages of a company's journey, the fair market value of common stock is very low. Instead of giving employees options, give them stock. The typical way to do that is to issue Restricted Stock Awards (RSAs). The employee will pay some taxes but probably very little, and they will need to do a bit more paperwork (e.g. file an 83(b) election within 30 days of the grant). In return, they become a shareholder right away, and start the long-term capital gains clock.
Once a company's FMV grows (e.g. beyond $8-10M), it becomes less practical. E.g. a 0.5% grant vesting over 4 years at $8M FMV would represent a $10,000 gain each year. Pretty quickly, you need to revert to using options (ISOs).

Note that FMV for common stock != the value of your last round (as in the price of preferred stock). It often is much lower. </blockquote>
In most cases, you would be able to issue RSAs until you raise your first priced round (vs. SAFEs). This is in effect what founders do for themselves: issue shares at a very low price (usually par value) in the first days of the company. I am amazed that it is not more common for employees.
Issue early-exercisable ISOs until series C</h3>
Until the Series C, a company's shares are often QSBS eligible. This is an incredible tax benefit which may allow your employees to collect long term capital gains tax-free, provided they held stock for a minimum of 5 years.
Once RSAs are no longer practical, ISOs are usually the best option available. They do not cost employees anything on grant or upon vesting, and they give them the flexibility to exercise them when it makes sense for them.
Allowing early exercise on those options is simply a matter of offering maximum flexibility.
Issue double trigger, 10-year RSUs thereafter</h3>
After the Series C, two things often happen:

The company's valuation is so high that exercising options becomes very expensive</li>
The valuation also means you'll quickly run up against the $100k limit per year for ISOs (calculated as FMV * number of shares)</li>
The company's shares are no longer QSBS eligible</li> </ol>
At that point, a double trigger RSU is probably the best instrument for employees.
Educate employees on equity</h3>
Employees are often very naive about the mechanics of equity compensation. To avoid big surprises down the line, companies should educate them.
Bring in a tax advisor once a year to present to employees and host a Q&A. The provider will see it as a great lead gen opportunity, and the employees will appreciate the information.
There are companies who specialize in this kind of training, such as Equity FTW</a>. A simple place to start is Carta's excellent education site</a>.
Conditionally offer longer exercise windows</h3>
At the risk of stepping into controversy: I am less enthusiastic about a 10-year exercise windows than I once was. While in theory it is better than nothing for the employee, in practice employees rarely consider the difference between ISO and NSO carefully enough. For the company, it is burdensome at the liquidity event.
Instead, I recommend conditioning exercise window extension on tenure or some other factor. Carta for example extends the exercise window by one year for every year of vesting. Other companies extend the exercise window to five or seven years after two or three years of employment.
This, to me, strikes a good balance between freeing employees from the 3-month exercise window and keeping the amount of outstanding NSOs a bit tighter.
Offer liquidity when possible</h3>

Disclaimer: i've never participated in or run a tender offer. </blockquote>
One of the main constraints with startup equity is that it is illiquid. The obvious answer is to offer liquidity. Regular tender offers allow employees to take money off the table, or to exercise their options without needing to drain their savings.
Of course this is easier said than done. While secondary sales are common alongside rounds of financing, offering liquidity to all employees typically requires a tender offer which is a lot more complicated (and expensive). The main hurdles are:

Not every company can easily line up buyers in tender offers. Fundraising is hard enough as it is!</li>
Very few company have the cash to buy the shares themselves</li>
It takes a lot of time and effort.</li>
It generates large legal bills.</li> </ol>
As a result, tender offers are rare until the very late stage.
Still, they are becoming more common as businesses take longer and longer to IPO. Leading companies have embraced regular tender offers. Stripe, Databricks, OpenAI, Anthropic, Gusto, and Notion have all run tender offers in recent years. SpaceX runs them every six months.
Conclusion</h2>
This turned out quite a bit longer than I expected! The topic is complex, and matters of law and tax are subject to interpretation. Still it is clear to me that the current setup is not fit for purpose. In a world where startups stay private for decades, options that expire after 10 years with a three-month exercise window do not work.
The setup I described above is the best option for most startups, as I understand it. And it is what I would do if I started another company. But a true solution requires regulatory changes.
If it were up to me, I would adjust the exercise window for ISOs to 10 years. This would give employees greater latitude to leave and would greatly simplify the payout process in the acquisition of an earlier stage company. Raising the ISO limit a bit would be a good idea as well, $100k is not what it used to be.
At later stages, RSUs work well enough. You could even argue that the payroll tax liability that comes with them is a feature not a bug.
Have you seen interesting structures? Do you have ideas for better solutions? I'd love to hear from you!
Further Readings:

https://a16z.com/recommendations-for-startup-employee-option-plans/</li>
https://zachholman.com/posts/the-new-10-year-vesting-schedule</li>
https://dangelo.quora.com/10-Year-Exercise-Periods-Make-Sense?share=1</li>
https://www.benkuhn.net/clawback/</li>
https://carta.com/learn/equity/</li> </ul>

Challenging the two party system

2025-03-02T00:00:00+00:00

The United States has a two-party system implemented through closed primaries, first-past-the-post voting, and the rules of the electoral college. Critics argue that it is unrepresentative, leads to polarization, and disenfranchises voters. Ironically, the writers of the Constitution were hostile to political parties (something both Hamilton and Madison agreed on: they respectively warned of "the pestilential influence of party animosities"</a> and "the mischiefs of faction"</a>).
High-profile challenges to the status quo have focused on the presidency, with a recent example being the effort by No Labels to secure ballot access and recruit candidates for the 2024 US presidential elections. However, the presidency presents the greatest barriers for third-party candidates: unless a majority of electoral college votes go to one candidate, the decision falls to the House of Representatives. Most states ban faithless electors these days, so no coalition could emerge and a real three way race would end up decided by the House more often than not.
The House seems a much better target for a third-party effort, and the current moment feels especially ripe for it. House races are cheaper and more responsive to local politics. With widespread dissatisfaction toward both major parties, persistent congressional gridlock, and razor-thin majorities in the House, even a small bloc of third-party representatives could wield significant influence as tiebreakers on contentious legislation.
As far as I can tell, existing efforts focus on reforming election law to make the House more representative. Organizations such as Unite America</a>, Andrew Yang's Forward Party</a>, and Fix Our House</a> all advocate for similar solutions:

Ranked-Choice Voting</li>
Open Primaries</li>
Independent Redistricting Commissions</li> </ol>
These all strike me as good ideas, and given that they can be adopted by individual states without requiring a change in federal law, one can hope that they will gradually be enacted in more and more states. Alaska is leading the way since Ballot Measure 2</a> implemented a top-four open primary with ranked-choice general election in November 2020. It is too early to tell how it will work out, but I note that it has already led to interesting situations such as the mutual endorsement</a> of a Republican and Democrat for statewide office. Unite America has a longer report</a> for those who are interested.
But I do not think reform is strictly necessary for a third party to thrive in the House. States such as Maine, Vermont, Alaska, or Utah all have strong identities centered on independence. Major issues such as housing, marijuana legalization, and corporate subsidies do not neatly align with the Democratic/Republican divide and could be used as a wedge. A strategically focused and adequately funded campaign targeting just 3-4 House seats could immediately position a third party with substantial legislative leverage. They would need great candidates, people like Dan Osborn in Nebraska or Evan McMullin in Utah, and a carefully crafted program focused on local politics and a few broadly popular national proposals.
Longer term, a national third party could eventually influence presidential politics. Imagine a third party trading their endorsement for a cabinet appointment and support on some of their key issues? This is not science fiction: I'd argue this is exactly what happened between RFK Jr. and the Trump campaign. To me, this is a much more credible path to third party politics than a No Labels ticket.
Perhaps the right place to start is state legislative races, where some progress is already being made. The United Utah Party (Utah), Vermont Progressive Party (Vermont), Working Families Party (New York), Alaska Independence Party (Alaska), and the Independence Party of Minnesota (Minnesota) all have the ability to win local races in their respective states. They could then build on electoral victories and legislative achievements to field competitive candidates for the House.
For now, we are a long way from a credible third party emerging in the House. But I believe it is possible, and I will gladly support it if one emerges.

Domaine des 2 Ânes, Premiers Pas

2025-02-23T00:00:00+00:00

Eric Asimov, the nephew of Isaac Asimov and more importantly one of the wine critics at the New York Times, published a great list</a> of wines under $20 he recommends last month.
I bought a few of them, and ended up really liking the "Premier Pas" Carignan from Domaine des 2 Ânes. It's not going to win any awards, but it's a good bottle for your next date night or dinner party.

Experimenting with AI Coding

2025-02-22T00:00:00+00:00

I have been experimenting with AI coding, and am blown away by the results. Aider in particular is an incredible tool, and I cannot imagine ever going back to writing all my own code. The speed at which these tools are progressing is frightening.
As a firmware engineer, my coding habits are a little bit old school. I use vim</code> as my text editor of choice (well, Neovim these days), am very comfortable on the command line, and write mostly C code with some Python sprinkled on top.
After a quick review of the AI coding landscape, I elected to try Aider</a>, because I felt it matched my workflow best. Aider is an opensource command line tool for pair-programming with AI. It supports many different models under the hood, and even lets you mix and match models (more on this later).
As of this writing, here is my Aider invocation:
$ aider --architect --model o3-mini --reasoning-effort high --editor-model claude-3.5-sonnet</code></pre> Aider works like most AI coding tools: you enter some prompts, and it responds and edits your code. I have settled into the "architect" mode workflow: I tell it what I want, it proposes a change, I confirm that it is what I want and asks it to implement it. It feels - to me - like pair programming with a smart (and fast!) intern who knows the syntax and idioms of various languages much better than I do. However it regularly makes mistakes and if I give it too much to bite off in one prompt, it drowns. It has transformed my relationship to code. What might have taken me a day or two of work now takes just an hour. As a result, I reach for code a lot more often! I write little utilities all the time, because it costs me so litte. Just tonight, I wrote a little script to push updates to this blog over WebDAV (instead of manual drag and drop). It took me 15 minutes. What I have learned so far: The models do best with Python and (Java|Type)-script, though they are reasonable at all mainstream languages.</li> It seems to me like it does better with backend than frontend. Most likely because it is not very good at looking at a screenshot and associating it with the code to understand what needs to be changed.</li> You have to own the overall design of the software. Technology choices, high level architecture, ...</li> Anything you can do to create a fast feedback loop on code quality will greatly improve the workflow. Use a compiler, type-checker, linter (e.g. Ruff), and a lot of tests (which the AI itself can generate). Feed errors back into the model and ask it to fix them. After a few failed attempts, step in and guide the model.</li> One prompt should be one ~200 LOC pull request (not a hard rule, just a good thing to keep in mind). You cannot give it a super high level goal. Think about how you might solve the problem yourself, and add some hints to the prompt (e.g. Wrap this in a new class and instantiate it here).</li> Assume it only has memorized the API for the MOST COMMON libraries, just like you would another engineer! Have a conversation with the model about what 3rd party library they might use to solve the problem, validate it is what you would do (or suggest something else). Then drop some example code in your prompt. In aider, you can have these conversations with /ask</code></li> You can even pass it links to documentation. Aider will scrape them and add them to the prompt context.</li> </ol> </li> When done with one feature, clear the history so it does not pollute the next. Sometimes the model will use previous guidance you gave it to try to do the next task, but that is rarely what you want. In aider, you can do this with /clear</code></li> When debugging something, ask the model what data it needs! Often times I'll run ripgrep and give it the output. E.g. /ask What additional information might you need? Happy to run shell commands for you</code></li> </ol> </li> </ol>
Why are high end hotels such a crappy experience? 2025-02-21T00:00:00+00:00 A few years ago, I found myself staying at El Encanto in Santa Barbara to avoid the fires and smoke that were plaguing San Francisco. The hotel room itself was expensive (it goes for $800 per night on expedia as of this writing), but what struck me was how they were upselling me at every turn. Want to park your car? It is valet only, and $50 per night. The restaurant serves an unsurprising pasta dish for $46. A bottle of water in the room? that'll cost you... You get the idea. High end restaurants offer the opposite experience. More often than not, they offer a prix fixe menu which covers the whole experience. They even go out of their way to make you think you're getting additional treats at no charge. At Jean-Georges for example, an impeccably dressed waiter wheels out a cart of gourmandises at the end of the meal and encourages you to "have as many as you'd like". The contrast is stark! I imagine most people prefer the latter, although they might not want to pay for it. [Yes, I know all inclusive resorts exist. But why aren't they all?] Why the difference? My best guess is that most people search for and book hotels very differently than they search for and book restaurants, and the behavior drives the business model. Most people when they book a hotel are just looking for "any hotels in this area that meets these requirements", and decide based on price. Hotels are incentivized to keep their price low and make their money elsewhere. On the other hand, nobody comparison-shops restaurants. Instead, they are looking for a specific experience (which is why trendy reservations are impossible to get, while merely "good" restaurants can struggle to bring in customers).Memfault Raises Series B 2023-01-24T00:00:00+00:00 I am thrilled to announce that Memfault, the company Chris, Tyler and I co-founded in 2018, has closed a Series B funding round led by Stripes, with participation from Partech Partners, Uncork Capital, and the 5G Open Innovation Lab. When we started Memfault, our goal was to make building hardware products easier. Since then, we’ve been fortunate to work with incredible companies such as Bose, Lyft, Samsara, and Verkada, helping them adopt a proactive approach to device reliability engineering and ship category-defining products. Today, our platform serves millions of devices and processes over a billion events every month, and we’re just getting started. We believe that the Internet of Things has the power to change the way we live and work, but to realize this full potential, we need to ensure that connected products are reliable and performant. That’s why we’ve built our platform to help companies quickly and easily diagnose and fix issues in their connected products without the need for expensive and time-consuming physical debugging or technician visits. With this funding, we plan to expand our platform, adding new features and capabilities that will make it even easier for companies to build reliable connected products. We’ll also be growing the set of platforms our features are available on, so you can monitor your devices with Memfault no matter what operating system, chipset, and programming language you are using. Additionally, we’ll be investing in our team, hiring top talent to help us continue to innovate and push the boundaries of what’s possible with IoT. We are hiring</a>! We’re honored to have Stripes as our lead investor in this round, and we’re excited to work with them to help even more companies around the world build better-connected products. Together, we can help make the Internet of Things deliver on its promises for everyone. I want to take this opportunity to thank our customers, partners, and team for their continued support. I am excited about the future and can’t wait to see what we will achieve together! See also: Techcrunch</a>, Memfault</a>, Stripes</a>

Squaring the circle on employee equity

Issue double trigger, 10-year RSUs thereafter</h3> After the Series C, two things often happen:</p>

Challenging the two party system

Domaine des 2 Ânes, Premiers Pas

Experimenting with AI Coding

Why are high end hotels such a crappy experience?

Memfault Raises Series B