by Connor Jerzak, February 18, 2026
Most advice about “teams” is either platitude (“communicate!”) or pathology (postmortems after disaster). I want something closer to a field manual: what I’ve found that tends to work, what tends to break, and what you can do tonight that changes the trajectory.
This is mostly about research teams—students, RAs, collaborators—but some, I’m sure, generalizes to any setting where (i) attention is scarce, (ii) incentives are complex, (iii) progress is lumpy, and (iv) work is being done at the (knowledge) frontier.
1. Meaning is a coordination technology
The first job of a team lead is not “task allocation”. It is making the work legible and communicating the mission behind the work.
Mentees (and honestly, all of us!) are navigating a world optimized to fragment attention: classes, side projects, deadlines, jobs, social obligations, anxiety. In that environment, “do this because I said so” is not something that will bring out the best effort of a team. A project survives both the good weeks and the hard ones only if people can answer, quickly and in their own words:
- What are we trying to learn?
- Why does that matter outside this room?
- What will exist when we’re done that doesn’t exist today? What will be the unlock or the 10x?
If you can’t give people a reason, you’ll end up trying to substitute discipline for motivation. That works briefly, but fails to survive long stretches when a team might not get any concrete “wins” on the research front in terms of new discoveries or publications.
Heuristic: if someone misses meetings / goes quiet / drifts, don’t start with enforcement and don’t lose faith in them. Start by trying to explain to yourself, and then them — the underlying meaning behind what we’re doing. Most disengagement is not rebellion; it’s often a lack of clarity regarding the “why” of the pursuit.
2. Incentives: make them explicit, make them compatible
Team dynamics also can degrade when incentives are implicit, misunderstood, or when they’re different for different members of the team — some members might want a quick win before hiring season, whereas others might want to take the time required to develop a high-impact piece of work.
A student is not “working on your project”. They are trading time and attention for some bundle of returns. Typically, these returns would include:
- Money (RA wages),
- Skills (methods, writing, data work, taste),
- Signals (letters, authorship, evaluations),
- Intrinsic joy of discovery,
- Degree,
- Networks (who knows them; who will vouch for them).
The right question is not “Is a mentee motivated?” It’s: does the project pay them in the currency they’re here for?
When onboarding someone new, the core design problem is incentive compatibility: set expectations so that doing the right thing for the project is also the right thing for the student. This is where honest communication with the student is important. Sometimes, a team might not have a large grant to pull from, but are there other forms of compensation that might be just as valuable?
Operationally:
- Be concrete about what success yields (a preprint? a publication? a codebase? a letter with real content?).
- Avoid vague promises (“this might become a paper”) unless you also define what concrete and realistic intermediate wins look like.
- Try to group team members with similar time preferences on the same project — avoid situations where some group members require quick wins while others require big and often slower ones.
3. Default positive/open vibes
As a new team member or mentee, you may sometimes feel that more experienced members are too busy to help or don’t want to. In my experience, this is the wrong default attitude. Other people almost universally want to help. Cultivate: Default positivity, trust, helpfulness.
4. Sympathetic accountability
Every student has their own trajectory: constraints, confidence, home stressors, and variable prior training. Pretending otherwise can lead to uniform expectations that can frustrate those coming from very different (technical) backgrounds.
But sympathy without standards is also corrosive. The stable equilibrium is:
- a baseline model of what the person is dealing with, and
- clear expectations that the person can improve.
This is less “be nice” than it is “don’t hallucinate motives.” If you misread struggle as laziness, you might teach the student to hide or avoid confronting problems that can actually be addressed with an open mind.
A common failure mode: escalation can spiral when with someone feeling disrespected or misunderstood, then interpreting every subsequent interaction through that lens.
It is also important to rememeber that, if a mentee successfully completes the tasks suggested by the team lead — and the project fails — responsibility falls with the team lead. The team lead is ultimately responsible for the failure of the team in this case, and must self-reflect on how to better guide the team.
5. Experience is overrated; curiosity + agency are not
A surprising thing about training we’ve found: it can be easier to build good habits from scratch than to unlearn bad ones and relearn better ones.
So “low experience” in a method area is not necessarily a red flag. It can be an opportunity—if the person has:
- curiosity (they want to understand),
- agency (they move without being pushed),
- non-defeatism (they believe the work can be done).
Our team has benefited a lot from working with students who may have been inexperienced at first but are relentlessly iterative.
6. Low-stakes communication prevents spiraling
When a student or mentee self-isolates, negativity accelerates. This is a specific failure pattern: a student hits a snag, delays communicating, shame accumulates, delay grows, and soon they’re in a hole that feels too deep to climb out of. The “problem” becomes less the task and more the social cost of reappearing.
So we want a culture where meetings are not a tax or a show of “what the student has accomplished”. Meetings should be more like a tasting of our cooking mid-prep — do we need to add salt? More broth? They are a chance for the team to learn more about the deep insights stemming from “failed” experiments, or to use that apparent failure as the stimulus for a new insight. Actually, we teach our students that no experiment fails — every result is informative about some question or process, if we can only find it.
Norm to teach explicitly: “Bring along partial progress early. We optimize for cycle time, not perfection!”
7. Small teams are the default
By default, smaller teams are wiser.
Why? Because alignment is easier:
- Individual contributions are more visible,
- Credit is clearer,
- Incentives are easier to align,
- Discrepancies surface sooner,
- Coordination overhead is lower.
Large teams can work, but they require more formal processes to avoid diffusion of responsibility and quiet free-riding, which can generate resentment from the de facto team lead and even make them want to move on to the next opportunity.
Takeaway: if you’re not sure you need a bigger team, you don’t. The ideal team size is two to three.
8. The reinforcing loop (and how to get one)
Teams get “stuck” when progress feels invisible. Teams get strong when progress is socially contagious.
Frequent meetings + small wins can create a positive reinforcement cycle:
- One person makes a step forward,
- Others see momentum,
- Expectations ratchet up,
- Team starts to believe progress is not only possible, but normal.
If you want a good overall lab culture, engineer the ability of each team to learn about the wins of other groups, and to make such wins feel like a natural part of the group environment.
9. Don’t kill projects casually—but don’t worship them
Project termination is hard because projects are also training environments. A mentee might have invested months into a project that doesn’t seem to be progressing.
A project that yields nothing tangible often feels like a waste to students, and that demoralization is contagious across teams. Sometimes a “suboptimal” output (less prestigious, narrower venue) is still valuable here because it produces:
- a concrete artifact,
- skill development,
- a narrative of completion.
So: bias toward finishing—our current results might not be considered as “wins” according to our initial definition of victory, but how can we reframe the question so that these results actually are enough?
During the project, continually reflect:
- What deliverable counts as “done enough” (memo, replication package, submission at venue A),
- What would justify pivoting to a project reframe,
- What would justify pivoting to a new project (data failure, identification collapse, irrecoverable scope).
10. Break logjams with new perspective or task switching
When a mentee gets blocked, there are two high-leverage interventions:
1) Add new blood. A new collaborator can break harmonization problems—mismatched working styles, stale assumptions, narrow idea space.
2) Switch tasks temporarily. Changing the problem can restore novelty and generate synergies that unblock the original thread.
Stalled projects often aren’t “hard”; they’re trapped in a local minimum and need a new perspective to break the team out of its narrow view of the problem, or even what counts as success to begin with.
11. The underrated variable: fun
Research isn’t always fun. But a team that never experiences joy is a team that won’t survive the inevitable grind of working at the research frontier.
“Fun” here isn’t goofing off. It’s the pleasant exploration of ideas, the feeling of discovery, the sense that the work is alive and meaningful. That emotional tone is fuel for our work.
If you can keep a team curious and moving, all else in team life becomes easier.
I don’t see that it makes any point that someone in the Swedish Academy decides that this work is noble enough to receive a prize. I’ve already got the prize. The prize is the pleasure of finding the thing out, the kick in the discovery. — Richard P. Feynman [Video]
–