Prerequisite: Hamming questions


What is a Hamming circle for?  What does it do?

A Hamming circle is a tool/process for making some kind of progress on some large, significant bottleneck.

This is slightly vague, because Hamming circles are versatile.  Participants at CFAR workshops have brought, into their Hamming circles, all sorts of problems and questions and goals.  A sampling:

  • How to find a spouse
  • What to do about my deteriorating relationship with my teenage child
  • Is it possible for me, specifically, to have a meaningful impact on existential risk
  • Something just isn't right about my life
  • I need to secure $250,000 in seed funding for my startup
  • My partner and I keep having the same fight
  • Even though I know I need to exercise, I just keep not exercising
  • What if everything I'm doing is "fake" and I'm only doing it because I feel like I'm "supposed" to
  • I want to go on a trip.  It feels important to go on a trip.  But I don't know why, and I don't know where, and I don't know what this trip should be
  • I think I want to quit grad school
  • I'm expected to speak at my father's funeral and I have nothing but scathing, bitter, angry things to say

What happens in a Hamming circle is fairly similar to what happens in a pair debug, or even when you're just working through your problems in your own mind.  However, the problems tend to be larger or deeper or more confusing or intractable, with the hope being that people will bring their most pressing bottleneck.

(Though bringing something smaller is fine; there shouldn't be moral pressure to shoulder a heavier problem than one feels ready to handle.)

The idea is that, for the duration of the circle, instead of having access to just one brain's power, each participant will instead have 3-5x their usual working memory, 3-5x their usual wisdom and life experiences, 3-5x their usual perspective or field-of-view, 3-5x idea generators or problem solving strategies, etc.

Summary description

Okay, but what is it?

In short: you and 2-4 other people will sit down together, and spend approximately 20 minutes focusing on a single person and their Hamming problem.  Then you'll take a short break, and reconvene to do it again with the next person, and the next, and the next.


You don't need much for a Hamming circle, but the things you do need are fairly non-negotiable.

  1. Time.  It's important for each person's "turn" in the circle to be at least 20 minutes (so that they have sufficient time and space to properly inhabit their problem, and aren't rushing or skimming).  It's also important for those turns to last no more than 40-50 minutes (because a too-open-ended atmosphere leads to meander and empirically results in less actual progress).  It's also good to have flex time for people to take snack breaks or bathroom breaks, and to have the opportunity to give someone five or ten extra minutes if they need it.  

    Thus, a three-person Hamming circle should allot about 90 minutes, and a four-person one about 130 minutes.  If you go up to five people, it's best to make turns shorter rather than adding another 20-40 minutes (so a five-person Hamming circle should still only be maybe 140 minutes long).
  2. People.  The ideal Hamming circle contains four people; stretching to three or five works okay but having two or six people changes the dynamic a lot.  It's important that those people all have some degree of mutual trust and mutual fellow-feeling; at CFAR workshops, Hamming circles were on the next-to-last evening in part so that participants would have time to get to know one another a little bit.

    (In principle, the simple trade of "I'll give you my time and attention, in exchange for you giving me yours" should work even with strangers, but in practice it's impossible for most people to be sufficiently open and vulnerable if they do not at least a little bit know and trust the other people in the circle.)
  3. Atmosphere.  The atmosphere of a good Hamming circle is like a long, warm embrace. It's important to be physically comfortable, in a place with non-horrible lighting and non-horrible sound.  It helps a lot if everyone is low to the ground, and it helps a lot if people are physically close.  Pillows and blankets are strongly recommended, and it's not a good idea to have a table in between the participants, or to have two Hamming circles taking place within earshot.  Bring coffee, hot chocolate, water, snacks, maybe a little bit of alcohol.  The feel you're shooting for is a late-night conversation among friends around the campfire.
  4. Problems.  People do not have to have an absolutely clear sense of their problem, or to have made a final call about which problem, but they must absolutely have a reason to be there, something that they themselves want the circle's help with. Spectating, or one-way Hamming where you only help and do not yourself get helped, is worse than it seems at first blush; it does something negative and somewhat corrodes the (for lack of a better word) "spell."  It creates distance, where what the circle needs is intimacy.

    (This is not to say that you shouldn't be flexible; sometimes crises arise and for various reasons it may be that one person ends up forsaking their turn to make space for someone else, or something.  That's okay.  It's just not good to plan on not taking your turn, and to show up with the intention of ... wearing clothes while everyone else is naked?)

The flow

It's usually best to begin a Hamming circle with some sort of easing-into-the-mood; at CFAR workshops this was accomplished by having the whole group gather to intro the activity before breaking off into smaller groups.  Having someone who's in charge, who knows what's going to happen and can speak calmly and softly and sort of bring the speed and temperature down is helpful.  

Once your group of (ideally) four has found a place that's quiet and isolated and physically comfortable, it's a good idea to take 1-3 minutes of quiet contemplation, where people can take a dozen deep breaths or close their eyes or do some brief Focusing or similar.  

Then (even if everyone present has already been in a Hamming circle before!) it helps to have someone explicitly set context, and to remind the group okay, we're here to try to make progress on our most pressing bottlenecks; this can be scary and difficult; we're here for each other; the trade we're making is showing up for others in exchange for them showing up for us.

After that, the group should organically choose who goes first, based on who feels most ready or most called or similar.

Each turn should have an official time-tracker; someone who can e.g. give a five-minute warning and then gently check in the last minute whether the group needs more time. That time-tracker should make sure, if they set a phone alarm, that the sound the alarm makes when it goes off is gentle and non-jarring, because it could easily come at an emotionally intense moment.

Participants should try to spend no more than five minutes setting context and explaining the problem and getting the group up to speed; it's very easy to accidentally spend 75% of your time on explaining, and end up getting very little in the way of reflection or help.  

(There is an exception to this general principle; more on that in the next section.)

After twenty-ish minutes have passed, and the person's turn ends, it's usually good to check in whether they need any aftercare, and for the group to do a (brief) moment of gratitude or hug it out or similar.  Then it's best to take a 5-10min break, to stretch legs and catch breath and grab food or use the bathroom, before returning for the next turn.

OODA Loops

(See the writeup of OODA loops for more detail)

You could model human behavior as looping repeatedly through four steps: Observe, Orient, Decide, Act.

(Another framing of these same steps is notice, orient, choose, execute.)

In generic applied rationality activities like pair debugging, people often find themselves distributed roughly evenly across these four steps.  Some bugs are about execution, others about making decisions, others about orienting, etc.  

In Hamming circles, however, there is a strong bias toward the first two.  The largest, stickiest, and most pressing problems often have their roots in unexpected places, or are entangled with all sorts of other habits or relationships or what-have-you that do not seem immediately connected.  It's often more helpful, in the Hamming circle context, to focus on understanding what's even going on than to try to leap aggressively toward solutions.

Thus, the "only spend five minutes getting the group up to speed" might not make sense, if the whole theme of your Hamming circle is "I don't even know what's going on, precisely?  I just know that something isn't working."  

Sometimes, the value of the Hamming circle is in laying out all of your observations on the table, having space to finally say what you've never quite managed to say, having other people (gently) poke and prod and make connections and draw out your reactions, etc. etc.

For instance, in the last of the examples given above (an individual who's expected to speak at a funeral but has nothing kind to say), it would be easy to snap into a particular frame ("oh, okay, let's help you brainstorm nice things!" or "oh, okay, let's get you out of this obligation!").  More likely, though, what this person needs is help orienting to the problem—figuring out what the problem even is.  If the group presupposes "ah, this is an issue of your family members not respecting your boundaries," then the circle is likely to be unhelpful or possibly even counterproductive.

Miscellaneous wisdom

  • Give yourself permission to not go in too deeply/wade out into treacherous waters/really drive yourself into a hole.  Hamming circles can be an excellent place to find support and lean on your friends and colleagues, but there is still a limit, and it is not virtuous to drive yourself into crisis.
  • As best you can, try not to worry about the other participants' experience, when it is your turn to be the focus of the circle.  Do not try to entertain them, do not try to make sure they're having a good time, do not sacrifice your own goals for the sake of everyday social niceties.  The Hamming circle is a special context, deliberately constructed such that you can set aside some of the social duties that you need to do in ordinary interaction.  Feel free to interrupt, or to redirect, or to make blunt requests of people; try to use the time in whatever way feels actually useful to you. Remember that, when it's their turn, they will do the same, and you will return the favor.
  • Try to avoid thinking in terms of solving problems with finality.  Be open to that possibility, if it arises, but don't shoot for it, as a target.  The goal of a Hamming circle is usually more about finding threads to pull, or increasing the surface area/grabbable parts of the problem.  Success is measured more in terms of clarity and understanding than in an ordinary pair debug or similar.
  • In general, don't do Hamming circles super frequently; they tend to lose their power/have a half-life of a couple of months.  Most CFAR staff and participants found that the value of a Hamming circle every 6-18 months was quite high, and the value of doing three in a six month period dropped off fairly steeply.

Call for crowdsourcing

By this point, there have easily been a thousand participants in official, by-that-name Hamming circles, and probably many many others who have taken part in other activities that either evolved from Hamming circles or are convergent evolution from elsewhere.

Please leave your own tips, suggestions, wisdom, and anecdotes below—anything that you think would help others avoid important pitfalls, or achieve particularly good outcomes.

New Comment
6 comments, sorted by Click to highlight new comments since: Today at 8:51 PM

Please leave your own tips, suggestions, wisdom, and anecdotes below—anything that you think would help others avoid important pitfalls, or achieve particularly good outcomes.

My old group debugging guidelines are more oriented towards regular (e.g. weekly) debugging sessions than Hamming circles, but still have a fair bit of overlap.

(Note that this version edited out the "Look for positive and negative reinforcers in the environment" approach which you can find in the original behind the link; I currently think that frame is pointing to something true, but also likely to be more misleading than beneficial.)


Only bring up something if you actually want a solution for it. Everyone has times when they just need to vent and want sympathy rather than solutions, but debugging circles aren't the place for that. This doesn't mean that you would need to accept any suggestion that the others bring up, but it does mean that that you should be open to others offering suggestions in general. Once the session ends, you're free to just ignore and forget anything that didn't seem to make sense to you.

Be courteous of others and their time. Do your best to make sure that everyone, both you and the others, get a roughly equal share of the group's attention. If you have lots of problems in your life, don't dump all of them on the group at once, but rather focus on one or a small set of related ones. If it starts looking like the discussion has gotten stuck on one person's issues for an extended time and the others might not have a chance to have their issues discussed, gently but firmly suggest moving on to the next person. Try to be considerate enough to pass on your own turn early enough that someone else doesn't need to prompt you to do so.

If someone is undergoing a particularly difficult time or has a particularly important issue going on, it's alright to sometimes spend a disproportionate time on them: but you should try to avoid being that person each time.

But have a fair respect of your own time, as well. The opposite also applies: if you genuinely feel that there's nothing in your life that needs discussing, you're free to cut your own turn short, but if you do it many occasions in a row, you're probably not taking full advantage of the group. If nothing else, you can always use the group to get an opinion on any assumptions behind your current plans. You have a right to get help from the group in return for helping others: stick to that right.

Don't proselytize your view. Maybe you're completely certain that the cause of the other person's problems is that they don't have cat ears as a part of their attire, which would totally fix everything if they just changed that. You're free to think that, but if they disagree with your suggestion, don't get stuck arguing but let it go.


Start by trying to understand what problem the person is trying to solve. "I've been trying to sign up for dance lessons but can never seem to get around it." One possible approach would be to immediately start offering ways for the person to sign up for dance lessons. Often a more fruitful one would be to first ask - why do you want to attend dance lessons? Maybe it turns out that the person doesn't actually care about learning to dance, but is feeling bad because their friend, a great dancer, always gets all the attention at parties. Then the actual problem is not "how to learn to dance" but "how to get other people to notice me". It's quite possible that not knowing to dance isn't actually the biggest issue there.

Test your understanding of the problem. When you're formulating an understanding of the problem, it can be useful to frequently verbalize it to the other person to make sure that you've understood correctly. "So you seem to be feeling bad because your partner just became the President of your country while you mostly spend time playing video games, is that right?"

A rule of thumb that I sometimes use is "do I feel like I understand this problem and its causes well enough that I could explain to a third person why this person wants to solve it and why they haven't been able to solve it yet?" If the answer is no, hold off proposing solutions and try asking more questions first.

Even "obvious" problems may benefit from questions. Someone once mentioned that they tend to often jump to being critical of others, which tends to be harmful. Here the causal mechanism seemed to be pretty obvious, but asking "how does it tend to be harmful" was still useful in bringing out details of the exact nature of the typical criticism and how people tended to react to that.

Look for trigger-action patterns. "I always end up being on the computer and wasting time and then feeling bad." What specific things on the computer act as time-wasters, and how exactly does the person end up doing those things? Maybe they often feel bored or anxious, which causes them to open Facebook, which causes them to get lost in a maze of discussions and links. Would there be a way to either remove the anxiety, or find a new action to carry out when anxious? Which one would be easier?

Be specific about the causes of emotional reactions. "My boss is so full of himself, it drives me nuts." Exactly how does the full-of-himself-ness manifest? If the exact behavior is "he often interrupts", maybe something could be done about that thing in particular. Best case: the boss comes from a conversational culture where interrupting is normal, and hasn't even realized that someone would consider it rude - but this would have been impossible for the others to suggest if the problem description would only have been on the level of "he's so full of himself".

This is also a useful technique for reducing your own annoyance at others, even if it was just something you did in your head. "I'm getting frustrated now because that person is talking really loudly and I would like to read." Breaking down an atomic "AAAAAGH I'M SO FRUSTRATED" into a "I'm feeling [specific emotion] because [specific cause] and [that violates my desire/need to something]" is not only useful for debugging, it can also relieve the frustration by itself.

Assume that problems won't fix themselves. In one session, someone says they intend to implement some change for next week's meeting. In the next session, they say, "yeah, that plan didn't really work out, but I was kinda busy and distracted this week. I'm going to try harder."

Chances are, if they were busy and distracted this week, they're likely to be busy and distracted the next week, too. "I'm going to try harder" often translates either as "I don't actually care about solving this problem but want to give the impression that I do", or alternatively, "I don't actually know how to fix this but I'm going to try again the same way, in the hopes of magically getting a different result now". Assuming that the person really does want to solve their problem, try to figure out exactly what went wrong and how it could be avoided in the future.

Ask, "is there a more general problem here?" Someone wants to cut down on the amount of money that they spend on fast food. One day when they're coming home from work they walk past a hamburger place, are tempted by the advertisements, and go there to eat. This happens several times.

The specific problem in this case would be "I always end up eating at the Burger King on the 27th street on my way home". The more general form of the problem might be something like "each time I walk past a fast food place when I'm hungry, I end up eating there". General solutions might be "pick a route that allows you to avoid seeing fast food places when you're hungry" and "make sure to carry something with you that allows you stave off the worst of the hunger until you're home".

FocusingSomeone is having difficulties deciding whether to try to solve a problem or whether to accept its consequences and let it be. One approach would be to have them verbalize all the reasons why the unsolved issue bothers them, and then say out loud, "having considered all of these consequences of the problem, I find that they're acceptable and it's better to just let this be". Does saying that feel right to them, or does something about it feel wrong? What if they were to say, "having considered all of these consequences of the problem, I find that they're unacceptable and I want to solve the problem", instead? Would that feel right or wrong?

Quick MurphyjitsuAfter you've come up with a plan, it may be useful to have the other person do a quick Murphyjitsu on it. How surprised would they be if this plan failed? If not particularly, is there any obvious failure mode that comes to mind and which could be fixed?

Check that the person remembers something actionable. Sometimes discussion may suggest some actionable things, then drift to e.g. more general discussion of the problem which doesn't provide as many concrete suggestions. If this happens, make sure that the person whose problems are being debugged still remembers the actionable suggestions they got earlier on.

This is a great write-up!

re: "As best you can, try not to worry about the other participants' experience, when it is your turn to be the focus of the circle...." 
I think it's pretty important to hit this point strongly in the initial context setting part, ala "Then (even if everyone present has already been in a Hamming circle before!) it helps to have someone explicitly set context, and to remind the group".

Even though you had the paragraph referenced above and the paragraph with "The idea is that, for the duration of the circle, instead of having access to just one brain's power...." I want to hit even harder that everyone in the hamming circle, when it's not their turn to be in the focus, are genuinely there to be of service to the primary individual. There's an important leaving of ego behind, an important checking at the door any sort of politeness, a handshake that says, "No really. We want to hear what you have to say, fully truthfully vulnerably. We want to give you our full undivided attention here. We're going to help you in whatever way you want to utilize the help we have on offer." This both serves to encourage the one in focus to relax / open up / actually do the thing, and it encourages the ones in the support role to focus on why they're there vs eg "doing it right" or "not saying the wrong thing". 

While people show up with good intention, they're not always able to follow the format of Hamming Circles. Sometimes the one in the center isn't willing to interrupt. Sometimes the ones supporting don't notice that they're not synced with the main individual. As the facilitator, I make sure to be tracking this. An example of that in practice: if it's Bob's time in the center and Alice has been talking for 3 minutes straight, I might interject with, "Hey, sorry to interrupt, but I want to double-check: Bob, is this the thread you want to be on?" Clumsily done, it can make things worse than better, but I do prefer installing these kinds of safety rails as the facilitator.

Also while "Prerequisite: Hamming questions" is at the top here, I want to emphasize explicitly that doing the Hamming questions or something like it is essential to successfully doing a hamming circle.

I noticed that it's been 3 months since this was posted. When can we expect more CFAR content?

I noticed that it's been 3 months since this was posted. When can we expect more CFAR content?

I’ve run Hamming circles within CFAR contexts a few times, and once outside. Tips from outside:

Timing can be tricky here! If you do 4x 20m with breaks, and you’re doing this in an evening, then by the time you get to the last person, people might be tired.

Especially so if you started with the Hamming Questions worksheet exercise (link as prereq at top of post).

I think next time I would drop to 15 each, and keep the worksheet.

No particular tips about Hamming Circles proper—I've run them a couple times but don't feel like I grokked how to run them well—but I'll put out that I've had some success with running longer events oriented towards Hamming Problems, shaped more like "here's 5 hours, broken into 25min pomodoros where you focus on making actual tangible progress towards something that's stuck, then during 5min breaks check in with a partner about how your focus is going".

Which on reflection is actually very similar to how the online goal-crafting intensives I've been running for years are structured, except with coaches instead of a buddy.