Choosing a Co-Founder: What Saju Can Tell You

A compatible chart cannot verify a compatible co-founder — real work, under real deadlines, tells you what a reading never could.

Saju cannot verify that a co-founder is honest, competent or committed. At most, it can suggest questions about pace, control, conflict, standards and uncertainty. Choose a co-founder through observed work, references, explicit agreements and legal due diligence, not chart compatibility alone. The strongest use of a compatibility reading is improving the conversation before commitment — it should surface assumptions to test, not pronounce two people destined to succeed or fail.

Two founders, one open question

Call them Founder A and Founder B. They'd known each other for a year, liked working together, and were now deciding whether to actually incorporate. "Are we compatible?" was the question they kept circling, and it was too vague to answer, because a real co-founder relationship is actually several agreements running at once: which problem they were committing to solve, how much time and money each of them would put in, who would decide when they disagreed, which functions each of them would own, how equity would vest, and what would happen if one of them left. Two people can enjoy each other and still be poor business partners; they can also have very different working styles and build something strong together, if authority and conflict are designed for rather than assumed away.

A Saju reading one of them had gotten for fun raised a few genuine questions worth checking: whether one of them needed rapid movement while the other needed proof before acting, who tended to centralize decisions under pressure and who withdrew, whether one preserved options while the other made early commitments. None of that was a fixed trait certified by birth data, and treating "their elements clash" as an explanation would have told them nothing. Converted into a real question — in the last two deadlines, had one of them changed scope without agreement while the other avoided challenging it? — it became something they could actually check against what had already happened between them.

What the chart couldn't tell either of them was whether the other one told the truth, could actually build or sell the product, would work full time, had undisclosed debt, treated employees well, owned the IP they claimed to, or would stay committed after real conflict. Those needed references, documents, observed behavior and, eventually, a lawyer. Saju's personality and predictive claims have not been scientifically validated, and neither a favorable reading nor an unfavorable one should be allowed to replace a difficult, factual conversation.

The trial that told them more than either reading

The most useful test A and B ran was completing real work together before either of them signed anything. Before incorporating or making a large equity commitment, that meant a time-bounded trial with:

one real customer or user problem
a defined deliverable
clear ownership
a deadline
a small budget
at least one moment of disagreement
a retrospective at the end

They didn't manufacture the conflict. The work itself was substantial enough that a real disagreement showed up on its own, over how much to promise a customer before the feature was actually built. How each of them behaved when the work got ambiguous, tedious and briefly adversarial told them more than the trial's actual output did.

What the trial actually let them check

Running that trial gave them a way to compare things a compatibility score never could: whether stated commitment (hours, start dates, financial needs, competing obligations) matched reality; whether their individual work samples and references held up, and which essential functions still had no clear owner even with both of them performing well; whether promises and deadlines arrived without repeated prompting; how each of them updated a belief after being shown they'd gotten something wrong, rather than performing certainty about it; how an actual disagreement got named, whether decision rights were clear inside it, and whether trust got repaired afterward; and, hardest to fake, what each of them would actually sacrifice when growth, quality, employee welfare and runway pulled in different directions. A chart could inspire questions in each of these areas. It never could have supplied the evidence the trial did.

Turning "we decide together" into something specific

"We decide everything together" tends to work only while decisions are easy. For each major domain (product, engineering, sales, hiring, finance, fundraising), A and B eventually had to write down who recommends, who must be consulted, who makes the final call, which decisions need unanimous consent, and what happens in a deadlock. Authority was allowed to change as the company changed, as long as the agreement said up front how that review would happen. When one of them got cast as "the decisive one" and the other as "the analytical one," they were careful not to let those labels harden into a permanent hierarchy — domain competence and the actual stakes of each decision mattered more than the label.

Equity got discussed only after the work itself had been discussed. It represents future contribution and risk, not just the idea that existed before the company started. Y Combinator's general guidance leans toward equal or close-to-equal splits for genuinely committed co-founders, with vesting (commonly four years with a one-year cliff) rather than an immediate, fully-owned grant. That's general startup guidance, not legal advice, and local law, tax treatment, prior IP and cash investment all change the specifics enough that a qualified lawyer and tax adviser were worth the cost before anything got signed, covering percentages, salary timing, IP assignment, departure terms, board composition and voting, future dilution, and expenses or founder loans. If any of those topics felt too awkward to raise, that was itself useful information: the partnership wasn't ready for equity yet.

Imagining the failure before it happens

A and B also ran a pre-mortem: imagine the partnership has failed eighteen months from now, each person privately writes the five most likely causes, then compare lists. The overlap mattered more than either list alone. Unequal commitment, avoided conflict, incompatible risk tolerance, unclear product authority, resentment about equity, personal financial pressure, different definitions of success, and poor communication around bad news are the causes that show up most often across founder teams generally. For each one that made their own list, they wrote down an early warning sign and an agreed response in advance, which turns compatibility from a feeling into an operating design.

Red flags no compatibility score should override

Regardless of what any reading says, the partnership should pause if either person notices:

inconsistent accounts of prior work
refusal to provide relevant references
hidden conflicts of interest
pressure to sign before review
contempt during disagreement
repeated broken commitments
unwillingness to discuss vesting or departure
demands that the chart settle a business dispute

These are behavioral and contractual risks, and they should be treated as such regardless of how the compatibility reading came out.

A co-founder decision brief

Before committing, complete this together:

Our twelve-month objective is ___.
Founder A owns ___; Founder B owns ___.
Decisions requiring both founders are ___.
Our deadlock process is ___.
Our time and cash commitments are ___.
Our vesting and departure terms are ___.
The strongest evidence that we work well is ___.
The strongest evidence against the partnership is ___.
The pattern most likely to damage trust is ___.
We will test it by ___ before signing ___ on ___ date.

Have legal counsel convert the relevant decisions into enforceable documents — a private worksheet is not a founders' agreement.