This is a written second opinion on one working relationship that keeps failing the same way. Not a second opinion on someone else's advice. On the read you already have. One situation, one counterpart, your side only.
The fee is $2,000, flat. It buys a three to four page brief that ends in a prediction with a date on it. You grade the prediction at thirty days, against what happened, not against how the brief read. If the forecast misses, the second read is free and the miss goes into the published record.
Nothing is paid until a case file has been reviewed and accepted. Some case files are declined. Both outcomes are shown on this page, in full.
You keep the brief and the graded appendix.
The read draws on an internal behavioral reference I maintain for this work. It is not named here and not taught, and it is not what you are buying. What you are buying is the brief and its grade.
Seven questions, ten to fifteen minutes, free. Reviewed within two business days.
Both are composites. Real submissions are confidential and always will be, so these are built from the shape the real ones take rather than from anyone's actual situation. Nobody described below exists.
They are here as a pair because either one alone proves very little. A sample of work anyone can produce. What is harder to produce is the second document, and the second document is the reason the first one means anything.
Before either of them, the arithmetic that separates them.
Every case file is checked for one thing before I take any money: can this situation produce a prediction you could actually check in thirty days.
That reduces to a number. How often the behavior happens, multiplied by thirty days.
| As described | Expected in 30 days |
My floor | Decision | |
|---|---|---|---|---|
| First case file | 9 times in 8 weeks | 4.8 | 4 | Taken |
| Second case file | 4 times a year | 0.33 | 4 | Declined |
Nine times in eight weeks is a bit over one a week. Four times a year is a third of one a month. You can check both sums yourself, which is the point. The decision is arithmetic, not judgment, and it happens before payment.
Below four, a quiet month tells you nothing. A prediction that something will stop, made about something that happens once a quarter, is a prediction that will look correct whatever happens. That is not a forecast. It is a coin landing the way it was always going to land.
Over the last eight weeks, your co-founder has verbally agreed to nine separate decisions during your weekly Monday leadership meetings, only to reopen every single one within 24 to 72 hours. The reopening occurs either through public posts in the engineering Slack channel or via direct messages to the specific engineer holding the project ticket. When you explicitly solicit objections at the close of each Monday meeting, he surfaces none. During your Thursday one-to-ones, when you point out that the topic was already settled, he states that the discussion had not happened properly. In one instance, he circulated a front-end job description that directly contradicted an infrastructure-first priority agreed upon the prior week, framing it as an assumed expansion rather than a reversal. You own all non-engineering functions; he owns engineering equally. No acute external pressure is present in the business, such as fundraising, departures, or reorganizations, and your next board meeting is ten weeks out. Decisions do not get formally overturned; instead, they stall through engineer confusion. To address this pattern, you have tried asking for objections more directly before closing meetings and pointing out prior agreements in your one-to-ones, which has produced surface agreement in the room followed by the identical offline reopening within one to three days.
Your automatic instinct when a decision gets reopened is to treat the initial in-room agreement as binding and the post-meeting communication as a deliberate maneuver to bypass your authority. When your co-founder posts in the engineering channel or messages a developer directly, you read it as a strategic power move aimed at taking ground where his authority is strongest. When he states that a topic was not discussed properly, you read it as a retroactive rewrite of history and respond by trying to re-establish the meeting record or calling out the pattern directly in your one-to-ones.
This move protects your operational authority and your need for predictable execution. By framing the issue as bad faith, you defend the validity of the leadership meeting as the sole binding governance venue and protect yourself from feeling outmaneuvered in channels where you have less direct oversight.
What sits underneath this dynamic is a misunderstanding of how his technical review happens. Verbal assent in a group setting does not mark the end of his evaluation. Agreement in the room occurs because the social structure of the meeting creates an expectation of closure, but his actual assessment of technical risk happens only after he has had time alone with the details. When he contacts an engineer or posts on Slack, the observable path points toward reducing technical risk at the point of work, rather than acquiring territory or undermining governance. He routes the concern to the person holding the ticket because that is where the technical risk lives.
Acting on the bad-faith read carries a high operational cost across every decision cycle. Responding by closing meetings more forcefully, soliciting objections more aggressively, or relitigating whether a decision was already settled raises the social friction of voicing concerns in the room. Raising that pressure increases the cost of objecting during the meeting without removing the post-meeting review that generates the offline behavior. The result is more surface agreement during the meeting, followed by the same post-meeting reversal. Furthermore, escalating the issue to an authority debate converts a simple procedural gap into a structural co-founder conflict ten weeks before a board meeting. Relitigating the meeting record shifts the conversation away from identifying what conditions surface concerns before work starts and toward an unresolvable dispute over whose memory of the meeting is accurate.
The interlocking dynamic between your two ways of working creates a self-reinforcing loop. When you solicit general objections at the close of a meeting, you request an immediate evaluation. The behavior observed is consistent with a pattern where technical risk evaluation occurs offline in solitude rather than in real time during a group discussion. Soliciting general objections in a group setting produces verbal agreement because the meeting format demands closure, while his technical evaluation remains incomplete at that moment.
Once the meeting concludes, private review runs the decision against technical realities, surfacing unvoiced operational risks. The post-meeting contacts, posting in Slack or messaging the assigned engineer, route the concern directly to the execution point. Bypassing a direct message to you in favor of the technical channel is consistent with a pattern where risk mitigation is directed at the work ticket rather than handled through executive alignment.
When you observe this bypass, you interpret it as an attack on agreed governance and respond by re-asserting the meeting record or calling out the reversal in your next one-to-one. That response increases the interpersonal friction associated with raising post-meeting concerns. When raising concerns after the fact leads to relitigating the record, the cost of surfacing objections offline rises without making in-room objection easier during the next meeting. Consequently, on the subsequent decision cycle, silence in the room occurs once again, followed by the same offline routing once technical risks register after the meeting ends.
This read runs against an internal behavioral reference maintained and not taught. The lock is not generated by personal intent or character traits, but by the direct structural collision between a real-time meeting close and an offline technical review that occurs after the meeting closes.
At the close of every Monday leadership meeting, update the decision log to mark agreed items as pending technical review, and state that project tickets will be generated on Wednesday morning after a 24-hour review window.
The obvious alternative is to demand that he surface all objections in the Monday meeting before anyone leaves the room. That alternative fails because it attempts to force an immediate evaluation from someone whose technical review occurs offline. Demanding immediate objections in the meeting increases pressure to give verbal assent, which guarantees that the actual risk assessment will continue to happen post-meeting and route around you. The recommended change is unilateral, requires no prior agreement to initiate, and modifies a meeting step you already perform.
By establishing a formal 24-hour buffer before engineering tickets are assigned, you align project rollout with his offline review timeline. Rather than attempting to eliminate his post-meeting review, this move provides a dedicated, structured window for technical notes to be added to the decision log before work begins on Wednesday.
Executing this change will feel uncomfortable during the first two cycles because it requires deliberately delaying ticket pickup by roughly 40 hours. However, building that brief pause into the schedule removes the friction that currently stalls execution mid-week.
The situation in the second one is not a small problem. Someone's co-founder is rewriting their contribution before it reaches the board, four times a year, and has been for two years. It is serious, it is real, and the person describing it deserves an answer.
The answer is that I cannot sell them anything they could check.
Thanks for sending this through. It is a real situation and the exact kind of work I handle, but I cannot build a forecast from what is here yet, so I would rather ask than accept and produce something you cannot check. As you describe it, the board pack rewrites happen quarterly, four times a year, which works out to about 0.33 occurrences over a 30-day window. Because my floor to write a gradeable forecast is four occurrences in 30 days, I am going to pass for now. There is no charge for this review.
What I am missing is a higher-frequency behavior occurring inside the next month that I can track. You mentioned having a weekly one-to-one. Does any related version of this pattern turn up in those regular meetings?
The brief ends in a single prediction with a date, which you grade at 30 days. To write a forecast that can actually be wrong, I need a higher rate to measure against and enough qualifying occasions in the coming month where the behavior can resolve. Without those, I would be writing something that sounds right but cannot be checked, which is what this process is built to avoid. If you want to reply with details on whether this surfaces in your weekly meetings, I will look again, though I cannot promise it will clear the bar. If not, no problem at all.
Seven questions, ten to fifteen minutes, free. Reviewed within two business days.
Not the situation, and not the reading of it.
I could tell you what is happening in that second case file. The pattern is legible, and I would have views about it within an hour. What I could not do is tell you whether those views were right, because at a third of one occurrence a month there is nothing inside the next thirty days that would settle it either way.
So the honest description is narrow: I declined the gradeability, not the readability.
That distinction invites an obvious question, and it deserves a straight answer. If the reading is available, why not sell that on its own, for less?
Because a reading you cannot check is worth what readings are worth, and you can get one anywhere, from anyone, at any price. Plenty of people will tell you what is going on with your co-founder. Some of them will be right. None of them will tell you in advance how you would know.
The only thing here that is not already freely available is the part that can fail. Selling the rest without it would be selling the easy half and keeping the money.
Two things are worth noticing, and they are the two things I would want to check if I were you.
The first brief ends in a sentence that can be wrong. A date, a count, a named behavior, and a stated condition under which the read was mistaken. You grade it at thirty days. If it is wrong, the second read is free, and it goes into the published record as a miss.
The second one turned down money on arithmetic you can verify. That is the part that is difficult to fake, and it is why the first one is worth reading. Anyone can show you their best work. The gate is what makes the best work mean something.
Every engagement ends in a sentence that can be wrong.
The forecast names a behavior, a count, and a date inside the next thirty days, and it states the condition under which the read was mistaken. At day thirty you grade it against what happened, not against whether the brief felt right when you read it.
If it missed, two things follow. The first is a second read at no charge, claimed within thirty days of the check: a new brief, from the same account plus whatever the thirty days produced, ending in a new forecast on a new date. It does not include a second session unless the situation itself has changed. The second is that the miss goes into the published record next to everything else. Aggregate results will be published as engagements complete, misses included, and publication does not depend on the results being good. What gets published is the count, not the case. No situation, no counterpart, no client, ever.
No testimonials, no case studies, no client logos, no count of reads delivered. The case files above are composites and say so. If you are looking for evidence that this has already worked for someone like you, it does not exist yet, and this page is not going to imply otherwise.
What exists instead is structure.
The forecast can be wrong, in writing, on a date. Most advice about a failing working relationship is built so it can never be wrong: no date, no count, nothing to score. This is built the other way, and when it fails, the failure is recorded where you can see it.
The gate turns down money before it is paid. You watched it do that above, on arithmetic you can check with a calculator.
None of this is proof. It is the structure that makes proof possible, and hard to fake once it exists. The first engagements are what build that record, and until it exists there is nothing here to check but the structure itself.
The fee is $2,000, flat. It covers the review of your written account, the ninety minute session, the brief within three business days, the thirty day check, and the graded appendix. The case file and its review cost nothing, whichever way the decision goes.
There is no deadline on this and no counter on this page.
Not legal advice. Not mediation. Not therapy. Not ongoing work: the engagement ends at the thirty day check, unless a missed forecast earns the free second read. Not for a situation in live legal or HR proceedings.
And it is your side only. Nothing is collected about the other person.
One person: Johanna Farrimond. Physics and electrical engineering degrees. A master's in information assurance. Twenty years in IT, including classified networks. The reading is done by hand. No model writes any part of it.
Seven questions about the situation from your side. Ten to fifteen minutes. One of them asks how often the behavior happens, because that is the number the gate runs on. It is free, and it is reviewed within two business days.
Then one of two things. The engagement is accepted, and payment and the sequence above begin. Or it is declined, with a short note showing the arithmetic, the way the second case file above was declined. A decline costs you nothing.