pointless and uncaring

Ski 2025 (Alpental and Summit East)
2 March, 2025
- family
- photos
Comments
Habenero Red Cabbage Coleslaw
1 March, 2025
- recipes
Ingredients

Slaw
- ½ head red cabbage, thinly sliced
- 1 carrot, shredded
- ½ red onion, thinly sliced
- 2-3 slices pickled habanero, minced
- 2 tbsp pickling liquid (from habaneros)
Dressing
- 2 tbsp white vinegar
- 2 tbsp olive oil
- 1 tbsp Dijon mustard
- 1¼ tbsp honey
- 1 tsp celery seed
- ½ tsp garlic powder
- ½ tsp salt
- ¼ tsp black pepper
Instructions
1. Prepare the vegetables – Slice cabbage and onion, shred carrot, and mince habanero.
2. **Make the dressing
Comments
Chats vs Channels
17 January, 2025
- tech
I have been on Don Quixote type quest to get my coworkers to post useful discussions and try and mimimize chat use to 1:1 Basic suggestion if more than 3 people might be interesed in something it should be on a channel. But lets look at why i’ve failed.

Why devs moved from mail/chat -> slack/discord/teams

Despite my arguments below I think most software developers moved to slack/discord/teams because it had a higher signal to noise ratio to begin with. The smart people you moved first and found a quiet place they could talk to other smart people. Email because it is actually more open was a victim of its own success. Anyone hit you up and once they’re past the spam filter go to the top of your inbox. Worse “tooling” would often flood your inbox. Good tooling could be filtered (though your responsiblity) chatgpt o1 here

You’ll notice that channels that are two broad (too many users/topics) usually die off. Thats because of low signal.

Linking is subtly important. I can document a coversation on a pull request or put it in an email thread or github issue. Even if outside people can’t click it future insiders will benefit.

Why channels are better over chats.

Mostly discoverability. Channels can be linked and found latter via search. That search might suck now but should improve. The AI’s will eventually comsume your channels and produce some magic. Your chats will be lost to time.

With chats I have no way of disocvering it other than having a member drop me in. I can’t really get the info out other than sticking some bot in. If its a two person chat you have to start over and copy and paste.

Chats also will try and cross topics over time and become even more confused. (think about namign chats should it be named for the topic or the people in it).

Why people fall back to chats.

more to come
1. Knowing where the right channel is.
2. Notifications
3. Cross org work.
4. Meetigns goto chats not channels.
5. NO way to turn a chat into a thread.
Ways to make this better.

more to come

Give rights to share thread and channels across orges
Comments
Cancun
25 December, 2024
- family
- photos
Comments
Release Bets
20 December, 2024
- tech
The Problem

So you’re a big wig at some tech company and you want to make a big jump in quality. A common and reasonable sounding thing to would be to bring in some bar raisers. You don’t have enough people you trust to do it at the Pull Request level so you decide to have them review releases yet. (Continuous Deployment isn’t happening for …. reasons….). Now you have an interesting incentive problem.

Bar raisers don’t have much incentive to get the code out but they have a big incentive not to let a bad change out. Type I vs Type II errors. The bar raisers don’t have much signal about what releases/changes are dangerous or not so they’re gonna want to talk about each change. You’ll end up spending a constant amount of time on low risk releases eating up a lot of the review time budget. Higher risk release might get LESS discussion just because low risk items ate up available time. So how do you get a better signal?

Enter the bet

If you want to make a release you should give the bar raisers odds their release will cause an outage (defining outages is hard but lets assume we have one). If I’m really safe I might say this release only has a 1% chance of causing an outage (anyone that says 0% is a fool). Then the bar raiser now has some information and if they want to talk about the issue they need to ante a minimum amount of money to the bet (lets say $10). If after discussion the odds don’t change the bet is locked. If the odds change (a good thing new information being exposed!) then releaser forfeits the ante and the bar raiser can still take whatever odds that come out if they want. This both focuses the discussions using time most effectively and aligns incentives to bring out truth rather than block or override.

This works best for single change releases. If you do bulk you need PR reviewers to all give predictions and aggregate them which is harder but still doable. This is basically a prediction market. It works better the more people that can participate and the more liquid the market is. Ideally I’d be able to go bet against the release of some division far from me if I have reason to doubt. Insider Trading is a goodness here!
Comments