Background
Sau chairs the Bayesian methods working party. Things had been fine until a cluster of small signals started pointing the same direction: a London meeting you’d suggested had to be rearranged on the day, a follow-up question about moving it online went unanswered, and the next invite came two weeks out despite a prior agreement to move to weekly catchups. Around the same time, you’d sent him an update — via bayesianpca.com, the live-updating site you built to share your analysis — that may have landed as more polished and complete than what he’s been producing himself.
The site’s framing problem
The site’s nav header reads “Bayesian Methods Working Party” — not “notes” or “supplementary material” — with no credit to Sau or the group anywhere on the page, only your GitHub identity linked. The domain itself carries no claim on the working party’s name, but the on-page framing does. Separately, your name is not yet on the working party’s official member list — only Sau’s is — and he hasn’t corrected that.
The read
Weighing it honestly: this is more likely mild friction or quiet insecurity than active hostility. A chair’s standing in a volunteer structure rests almost entirely on being seen as the leading voice — a member’s independent output outpacing that, delivered under the group’s own name with no attribution, is a coherent, sufficient explanation for someone adding distance without saying anything. The uncorrected member list is the strongest single data point, being a near-zero-cost fix left undone. “Sabotage” of the conference is not supported by anything observed and was flagged as catastrophizing driven by how much this working party means personally, not by evidence — worth resisting that leap.
The plan
- Gate the site behind a login while it’s “under development,” tell Sau plainly and briefly just before the catchup in two weeks, and share credentials with him directly. This removes the public-comparison exposure without hiding anything from him specifically, and reframes the site from “already published, competing with the group’s output” to “in draft, shared with the chair as a courtesy.”
- Decide in advance when it goes public again — tied to the conference outcome or to the content feeling finished — so it doesn’t just quietly stay private and lose the value of being a durable, external asset.
- Raise the member list as a factual gap when the moment is right, not as a grievance.
- Give it space rather than chase it. Don’t send another update that reads as a further display until this resets.
Update — Aviva reached out to Sau (timeline revised)
Aviva invited Sau to present the working party’s work in York, offering to cover travel and accommodation — arranged before the recent downturn, not after it. That changes the causal read. The invitation went to Sau, not to you, and Sau almost certainly knows you’ve just joined Aviva. An unprompted approach from your own employer to sponsor him specifically, landing right as you two started working closely together, is not something he can easily read as coincidence — regardless of whether Somer is actually behind it, and there’s no way to confirm that and no reason to try asking.
Revised interpretation: the invitation likely isn’t unrelated good news, it’s a plausible contributing cause of the friction, and probably predates and runs deeper than the “wow” update. The threat to Sau isn’t only being technically outpaced — it’s the working party’s external legitimacy appearing to route through your employer and connections rather than through his own standing as chair. The update may have confirmed a fear the York invitation had already planted, not created it from scratch.
The leverage point
This does hand you real leverage — if Sau suspects further opportunities route through your orbit, he has a genuine incentive to keep the relationship functional. But leverage like this only works un-invoked. Referencing it, hinting at it, or even a warm note that nods at the Aviva/York connection would confirm the exact fear driving his withdrawal and turn a passive asymmetry into an active humiliation. Hold it silently. Let him raise York if he wants to; don’t pre-empt it, don’t congratulate him in any way that carries subtext. If contact is warranted, keep it plain and unrelated to the invitation entirely.
Update — telling Sau you’re not ready for the conference
Decision going into Friday’s catchup: tell Sau plainly that you won’t be ready to present at the conference. The real drivers are the recent death of your father, a new job, and roughly 20 hours a week already sunk into the working party on top of that — genuine capacity, not a reaction to his coldness. This has also always been flagged as a live risk rather than a firm commitment, so it lands as a foreseeable risk materialising, not a broken promise or a reversal.
Deliberately left out of the message: the catchups sliding from the agreed weekly cadence to fortnightly. True, and part of why the schedule slipped, but raising it in the same breath as the withdrawal risks reading as blame folded into an explanation — the same stacking problem that made the original bayesianpca.com update land badly. The personal circumstances are sufficient on their own and don’t invite a defensive response.
Aviva/York is not part of this move at all and stays out of the conversation entirely, consistent with the leverage point above.
Deliver it the way the guardrails prescribe elsewhere in this file — state it once, plainly, and stop. No over-justifying, no watching for his reaction and adding more to fill the silence.
The dual effect, and why Friday won’t give a clean read
Two things happen at once, and they partly offset:
- It removes the outshine threat — you’re not presenting anything, so the dynamic that plausibly drove his withdrawal (being outpaced by your independent output) resolves itself without ever needing to be named.
- It’s also a real letdown — he wanted the conference presentation, and losing your contribution likely lands extra work back on him. That’s a fair, legitimate reaction on his part, not a sign of anything darker, and doesn’t need managing.
Because both are live at once, expect a mixed reaction on Friday rather than a clean “he warmed up” or “he’s still cold” signal — treat it as a single data point, same caution as the keyboard moment, not a verdict. The letdown has a natural expiry once the conference has passed; the outshine relief, if real, resets the underlying dynamic more durably. Net trajectory should trend better from here even if Friday itself reads as flat or awkward.