koliber

I don't know where it stems from, but I think it's harder to fire people in union shops. In a non-union shop, you have to work with the person, their manager, and HR to see if you can correct someone's performance, and if that does not work, let them go. At a union shop, I imagine someone from the union will also be involved in this process, and it's hard to imagine how this process would be made easier through the involvement of another person, who happens to hold a bias.

You're forgiven, but not correct re: me being under-qualified to comment. I have interacted with union shops and non-union shops and have 25 years of professional experience, so please don't be dismissive.

I've worked with union shops as clients and I've hit a surprising amount of friction due to things which seemed union related. Things like "that's not my job and we need XYZ to do it" and what seemed like people slacking at their job from the outside. I once needed a desk moved a few feet which required a power strip to be plugged into a different wall outlet, and was told we need to wait for the electricians. Risking the wrath and fallout, I did it myself.

I've also seen this kind of behavior at non-union shops, but it was less explicitly stated and it was relatively easy to do something about it.

I'm sure you're correct about the quality of life at union shops.

I never worked at a union workplace. However, I’ve worked with underperforming colleagues who were could not be fired for whatever reason. It was miserable.

Please forgive me. I am not a Markov chain.

Hmmm. Didn't think about that.

In people there is a difference between unconscious hallucinations vs. intentional creativity. However, there might be situations where they're not distinguishable. In LLMs, it's hard to talk about intentionality.

I love where you took this.

If you tell me that trees are big, and trees are made of hard wood, I as a human am capable of asking whether trees feel pain. I don't think what you said is false and I am not familiar with computational theory to be able to debate it. People occasionally have novel creative insights that do not derive from past experience or knowledge, and that is what I think of when I think of creativity.

Humans created novel concepts like writing literally out of thin air. I like how the book "Guns, Steels, and Germs" describes that novel creative process and contrasts it via a disseminative derivation process.

Here's how I see it, but I'm not sure how valid my mental model is.

Imagine a source corpus that consists of:

Cows are big. Big animals are happy. Some other big animals include pigs, horses, and whales.

A Markov chain can only return verbatim combinations. So it might return "Cows are big animals" or "Are big animals happy".

An LLM can get a sense of meaning in these words and can return ideas expressed in the input corpus. So in this case it might say "Pigs and horses are happy". It's not limited to responding with verbatim sequences. It can be seen as a bit more creative.

However, LLMs will not be able to represent ideas that it has not encountered before. It won't be able to come up with truly novel concepts, or even ask questions about them. Humans (some at least) have that unbounded creativity that LLMs do not.

You're right that this happens. However, if I had to guess I would say that it's due to bad information / intel.

North Korea is a closed country with limited access to information (think before the internet). On the flip side, I don't think catching copyright violators would be high on the list of concerns of the people making these maps.

On the map of Europe, Poland has a strange nonexistent river flowing through it. It enters the Baltic Sea more or less where the Vistula river really does. But going south it mysteriously veers east into Belarus.

There is a difference between creating a summary and double-checking if a summary is correct. It's kind of like coding something, or doing a code review. It does save me time, but that might have to do with how I think and what I enjoy doing. Everyone is different.

I guess you’re right re chatbot for summaries. I was thinking about the use case and not the whole integrated product experience.

For example, for code gen I use agents like Claude Code, one-shot interfaces like Codex tasks, and chatbots like the generic ChatGPT. It depends on the task at hand, how much time I have, whether I am on the phone or on a laptop, and my mood. It’s all code gen though.

It depends on how much time I have, and how important the task is. I've been surprised and I've been disappointed.

One particular time I was wrestling with a CI/CD issue. I could not for the life of me figure it out. The logs were cryptic and there was a lot of them. In desperation I pasted the 10 or so pages of raw logs into ChatGPT and asked it to see if it can spot the problem. It have me three potential things to look at, and the first one was it.

By directing my attention it saved me a lot of time.

At the same time, I've seen it fail. I recently pasted about 10 meetings worth of conversation notes and asked it to summarize what one person said. It came back with garbage, mixed a bunch of things up, and in general did not come up with anything useful.

In some middle-of-the road cases, what you said mirrors my experience: we disagree what is notable and what is not. Still, this is a net positive. I take the stuff it gives me, discard the things I disagree on, and at least I have a partial summary. I generally check everything it spits out against the original and ask it to cite the original sources, so I don't end up with hallucinated facts. It's less time than writing up a summary myself, and it's the kind of work that I find more enjoyable than typing summaries.

Still, the hit to miss ration is good enough and the time savings on the hits are impressive so I continue to use it in various situations where I need a summary or I need it to direct my attention to something.

I hate what LLM spit out and would never accept the whole output verbatim.

I love how they occasionally come up with a turn of phrase, a thought path, or surprising perspective. I work with them iteratively to brainstorm, transform, and crate compose content that I incorporate into my own work.

Regarding spotting AI-generate content, I was once accused of posting AI-generated content where I bona-fide typed every single letter myself without as much as glancing at an LLM. People's confidence in spotting AI content will vary and err on fake-positives and fake-negatives too. My kids now think all CG movies are AI generated, even the ones that pre-date image and video gen. They're pretty sure it's AI though.

I sometimes get contracts, NDAs, or terms and conditions which normally I would automatically accept because they are low stakes and I don't have time to read them. At best I would skim them.

Now I pass them through an LLM and ask them to point out interesting, unconventional, or surprising things, and to summarize the document in a few bullet points. They're quite good at this, and I am can use what I discover later in my relationship with the counterparty in various ways.

I also use it to "summarize" a large log output and point out the interesting bits that are relevant to my inquiry.

Another use case is meeting notes. I use fireflies.ai for some of my meetings and the summaries are decent.

I guess summarization might not be the right word for all the cases, but it deals with going through the hay stack to find the needle.

I like this approach.

In finance you have three basic statements:

- balance sheet

- profit & loss

- cash flow

It seems like looking at net worth of a person is like only looking at the balance sheet. The mag wealth approach also takes potential cash flow into perspective, by estimating ability to raise cash, even if it is through loans. It does provide a more complete picture.

A few more seem to work as well, because I've used them and found them valuable

- human language translation

- summarization

- basic content generation

- spoken language transcription

Painting is such a great way to explain this. I've made many mistakes painting myself, and only came to appreciate the amount of extra work involved when I looked closely at how a pro does it.

Things that have caused me to screw up various paint jobs:

- not color matching white paint to my existing white paint, causing a visible color difference

- buying the wrong kind of brush for my paint

- buying cheap brushes and then needing to dig hairs out of the paint

- not priming, which caused me to have to do multiple layers to get good coverage

- buying cheap paint, which together with not priming, caused me to have to do multiple layers to get good coverage

- buying too little paint because I did not estimate carefully and then having to pause mid-job to go buy more paint and extending the total time to more than a week (because I did not have time to finish the job the next day)

- not buying and putting down floor protection which caused a lot of extra time in cleanup and never actually getting the paint out of the crevices in the floorboards

- not taping out the windows and doors which slowed down painting and required a lot of just-in-time cleanups

- taping out carelessly with poor-quality tape, causing paint to get under the tape, and defeating the purpose of taping

- not changing clothes before a tiny paint job and needing to throw away a very decent t-shirt because metal paint does not wash out well

- buying very expensive paint for a rough paint job in my garage because I did not spend a minute to think about the final effect, and wasting a bunch of money for a poor effect

To avoid each of these mistakes you have to spend time not-painting. It's time well spent.

I've tried doing a work breakdown, and watching a pro do it, it seems roughly 0.75 units of time for prep, 1 unit of time for painting, and 0.5 units of time for cleanup. I did not count the time for researching and buying, but once you add commute, it's easy for it to be 0.5 unit of time, depending on size of job. Roughly 1.5x-2x of not-painting to 1x of painting.

You can make many software-development analogies to painting.

If we're mature experience painters, all of these things not-painting things are still "doing things".

At the core, doing is paramount.

But doing the wrong thing, poorly, is also not doing the thing. To avoid that, you need to do some of the things that the post mentions.

That concludes my morning dose of Zen.

If you define “specifications” as a formal document with that title atop, then you’re right.

If you define it as a description of what needs to be built then by definition you can’t build what is not described.

User stories are a type of specification. “Build me a Facebook clone” is a type of specification. A well-refined Figma doc is a type of specification. They all specify what needs to be built at various levels of fidelity.

You need specifications.

It hasn't met my doctor.

You can't code without specifications, period. Specifications can have various forms but in ultimately define how your program should work.

The problem with what people call "Waterfall" is that there is an assumption that at some point you have a complete and correct spec and you code off of that.

A spec is never complete. Any methodology applied in a way that does not allow you to go back to revise and/or clarify specs will cause trouble. This was possible with waterfall and is more explicitly encouraged with various agile processes. How much it actually happens in practice differs regardless of how you name the methodology that you use.

That does put it into perspective. At the same time, I don't think a medium-sized container ship of pellets is lost each year.

They stated clearly in the article:

> This was our mistake, and we take full responsibility.

I wonder how much of the negative sentiment about this is from a knee jerk reaction and careless reading vs. thoughtful commentary.

Always round up (or down) to the nearest nickel regardless of whether someone pays card, cash, or SNAP. In effect, this would set the gross price as a multiple of 5 cents.

How would the apology need to be worded so that it does not get interpreted as a fake apology?

In terms of "downplaying" it seems like they are pretty concrete in sharing the blast radius. If less than 25% of users were affected, how else should they phrase this? They do say that this was data used for onboarding merchants that was on a system that was used in the past and is no longer used.

I am as annoyed by companies sugar coating responses, but here the response sounds refreshingly concrete and more genuine than most.

> Anywhere between 52,140 tonnes and 184,290 tonnes of pellets entered the environment in the EU in 2019, according to the European Commission, which proposed the regulations.

> "This is equivalent to between 2,100 and 7,300 trucks full of pellets per year," the Commission said.

How? These are unbelievable numbers. Did 2,100 - 7,300 trucks flip over and dump their nurdles into the environment? Are there steady nurdle-leaks which cause a percentage of them to escape? What percentage of nurdles escape rather than getting melted and molded into plastic end products?

> Beyond just heart disease & cancer taking you out entirely its: my eyesight is going, my hearing, every joint in my body could fail, my brain is slowing, etc.

Absolutely right. You can’t fix everything. But if you can only dedicate time, money, and attention to one thing, cardio health is probably highest impact for most people. This article makes that case. Also it makes the case that there are a few things that will have an outsized positive effect on cardio health and we’d be wise to focus on them.

People try to define the manager's job. There are many decent definitions, but one I really like is:

The essence of a manager's job is to push context down and pull information up. Here's a short article that explains it well: https://jacobian.org/2021/apr/19/the-fundamental-purpose-of-...

Everyone should be empowered to make as many decisions as possible. It gives people agency.

To make good decisions, you need context, knowledge, good judgement, and clear boundaries. The context gets built by managers who do a good job.

I've spent a decade of my career at 15Five where we've built a system of doing just that based on a weekly check in. When done well it worked marvelously.

Both the speed and cost problems can be solved by caching.

Each person gets their own cache. The format of the cache is a git repo tied to their sessionid. Each time a request is made it writes the code, html, CSS, and database to git and commits it. Over time you build more and more artifacts and fewer things need to get generated JIT. Should also help with stability.

Maybe LLMs will enable a new type of work in spreadsheets. Just like in coding we have PR reviews, with an LLM it should be possible to do a spreadsheet review. Ask the LLM to try to understand the intent and point out places where the spreadsheet deviates from the intent. Also ask the LLM to narrate the spreadsheet so it can be understood.