Meg Bolger

Thanks to @christsai for getting me to dig up my 2016 guide to single-threaded dinners. I think it mostly holds up and I would love suggestions. Hosting a Dinner (9/1/2016 edition; updated 9/11/2025) Here's my latest thinking about hosting dinners. I'm constantly revising this so please send me feedback. And I'm happy to be an emcee if I am free! Pre-work Research the social profiles of every attendee. You can use a service like FancyHands (or one of the AI agents) to put together a one-pager on attendees with 1) a photo 2) occupation and background (education, place of birth, etc) 3) latest few tweets and link to profile 4) something interesting about them (you can leave this open ended as a research task). You can do this a few days before and review it right before dinner. It is just for you and your staff. Seating You can decide on seating or not. Depends on the vibe of the dinner. Pre-introductions Just as everyone is sitting down, declare the subject of the evening, introduce the format of the evening and warn them that in a few minutes you are going to ask each of the attendees to introduce themselves and answer a question. What kind of question works well? Something that doesn't require deep domain knowledge but lets folks show off a little bit. I usually just ask something I'm curious about and use it as the starter for evening's discussion. One clever trick is to go meta: ask what question would you like answered about the subject of the evening. When you have a clear center of attention, you can ask something like “what question do you wish so-and-so would answers but are afraid to ask.” Rules of the road: Everything discussed is off the record; we'll have a single thread discussion until the main course is almost finished and then the conversation will dissolve into many conversations but until then you will moderate; at the end of the evening as dessert and coffee is being served, you'll ask everyone to share something they learned that evening. If you'd like, you can say that the best insight is going to win a prize. Introductions (after 5-10 minutes) Model how you want people to introduce themselves by offering a short introduction and then answering the question—you get to set the tone. Do you want it serious or light and funny? Then go around the room asking folks to introduce themselves. Depending on your goals, you can call each person out and say something flattering about them before you let them introduce themselves. (“And next, I'll let Mark introduce himself. He won't admit this but he's the main reason I started programming.”) Single-threaded conversation Imagine that you are a talk show host. Shush people who have side conversations. Move the conversation along to new topics when people look like they are getting bored. Declare an end to the single threaded conversation just before you think it is going to dissolve. Wrap Up As dessert comes out, ask people to share an insight or resource. If you offer a prize to the best insight, remind them of it before they share an insight. At the end, make a big show of giving something silly—another glass of wine or slice of pie. At the very end, ask if it is ok for you to send a mass email connecting everyone or share links and notes over WhatsApp.

Had a few closed door conversations with smart people concerned about AI this week. Here’s what I wish I told them: - there is an opportunity for prophets and critics but they’ve got spend time with the actual tools to get an intuition for what’s possible, rather than what you’d assume based on reading old science fiction. Frankenstein is a terrific book partly based on how concerned people were about electricity. It captures our fears about the nature of being human but didn’t help anyone really come up with better policies for dealing with electricity. I worry that a lot of AI critics are doing the same thing. - the tools are surprisingly accessible. I didn’t predict this but all of the big AI companies are spending an incredible amount of money to make their technology easy for people to use and sharing more about how they do it than I would have guessed. That’s not out of the goodness of their heart--they’ve got commercial reasons. But that is something you can use to your advantage! - the tools are also oddly cheap! For the price of flying out to San Francisco, you can run a whole series of experiments with the current tools to actually see what is possible, report on it, and then run another set of experiments. - the manifestos and policy papers are very good but everyone is still figuring out use cases so don’t spend all of your time worried about them. Remember ChatGPT’s success was a surprise. And the amazing NotebookLLM podcast I was showing off was subfeature of a subfeature. I think most at Google would admit they were surprised! - you can find out about the limitations of the tools by trying them out yourself rather than reading about the limitations. And if you play enough with them, you’ll start to build guesses about the ways they will and won’t get better. You’ll also see how hallucinations are both overhyped and still kind of under-studied. - I’m not a blind optimist. I would just love a world in which critics and prophets on the left and right illustrated their concerns with little demos and informed intuitions. There are lots of reasons to be worried, and lots of ideas that need to be reconsidered! But they should be grounded with the direction that the tools are going in right now, and the speculation should be on the second and third order effects of actual tools (which already are superhumanly intelligent in some ways), rather than an imaginary superhuman intelligences.

A fun chat with @thebenbergman about why I love investing in startups and what I’m looking for...