sunir

My frustration is the a/b split tests never seemed to net to anything in the long term even after confidence was reached. It made me question the entire process; but I understand the math so it’s confusing to me.

Maybe. Retargeting is unlikely to create novelty.

One of the most frustrating results I found is that A/B split tests often resolved into a winner within the sample size range we set; however if I left the split running over a longer period of time (eg a year) the difference would wash out.

I had retargeting in a 24 month split by accident and found it didn’t matter after all the cost in the long term. We could bend the conversion curve but not change the people who would convert.

And yes we did capture more revenue in the short term but over the long term the cost of the ads netted it all to zero or less than zero. And yes we turned off retreating after conversion. The result was customers who weren’t retargeted eventually bought anyway.

Has anyone else experienced the same?

There is already the shared source model.

What people call things is a marketing phenomenon. You can complain all you want.. and you should.

It’s not incoherent. It’s a rule that could be made and adhered to for a project.

It’s not possible for GPL projects to restrict the code this way, but the peripheral assets like trademarks, servers, conference slots, board seats, core contributor status could be restricted.

If you go down this path of creating rules that say people should pay to use open source, you’ll discover there are already laws in place sufficient to manage this situation. If a project uses these laws, then you’ll complain the project isn’t open source.

If you prefer the capricious nature of the politics of social shaming instead of the rules based system of laws and courts, I guess that is being consistent even if the actual process is very inconsistent and unpredictable.

If not, then it’s not clear to me how you’re taking a philosophical stance about open source if you’re demanding payment. Those ideas don’t work together.

Evan is amazing. I’m so grateful he keeps pushing an open internet forward.

I was thinking about the graphrag paper and prolog. I’d like to extract predicates. The source material will be inconsistent and contradictory and incomplete.

Using the clustering (community) model, an llm can summarize the opinions as a set of predicates which don’t have to agree and some general weight of how much people agree or disagree with them.

The predicates won’t be suitable for symbolic logic because the language will be loose. However an embedding model may be able to connect different symbols together.

Then you could attempt multiple runs through the database of predicates because there will be different opinions.

Then one could attempt to reason using these loosely stitched predicates. I don’t know how good the outcome would be.

I imagine this would be better in an interactive decision making tool where a human is evaluating the suggestions for the next step.

This could be better for planning than problem solving.

A more perfect union?

Just because democracy and capitalism were invented, didn’t change what humans knew worked beforehand.

The innovation is that you aren’t born into a given regime and you can switch regimes.

This made the system capable of any number of regimes from the freelancer to coops to communes to partnerships to private corporations to public corporations to crown corporations to charities to not-for-profit trade associations to your kids lemonade stand to nonincorporated affiliated entities to manage the local block yard sale to Uber to open source projects to family trusts to whatever and so on.

However to manage a market function to extract petroleum cost effectively, it seems a corporation at the scale of BP, as a hierarchical corporation is the system so far shown to be most efficient? I presume anyway.

Are your signs backwards?

There are a few ways. I built something similar huckai.com on top of vectara.com. They have open sourced their versions https://github.com/vectara

You can also do this on AWS now fairly easily. https://medium.com/data-reply-it-datatech/how-to-build-a-cus...

The lablab.ai Discord community is a pretty good place to learn how this product category is evolving.

I worked on an incredible C++ codebase back in the dot.com. We built a mobile SVG renderer and packed into 128kb on a Palm 3 and similar code sizes on Blackberries, Symbian (the worst OS ever), Fujitsu phones, and whatever you threw at us.

I built a few systems. The DOM, a build system that could cross compile to a completely new target phone while the phone executives were at lunch with our sales team (that’s great demo!), and an optimizing build step in Perl that rewrote the C++ to minimize bytes, cycles and joules of energy.

The rendering engine was built by a genius. It was the most efficient and accurate vector graphics renderer I have ever seen.

But of course no one could understand how this system worked.

I mean, Perl that transpiles C++ into more optimized C++ just is abusive. I was not popular with the senior devs.

The main coders including me were coop students who worked furiously at all hours for the fun of it.

But of course the product had no revenue. The wiser devs probably knew this and didn’t care.

Nevertheless, we eventually were let go during the dot.com bust and the remaining main engineers had to throw the code out and try to rewrite it in Java. That didn’t work because in that era Java was too slow to fit on a phone. Or maybe the market was dead.

I learnt a solid lesson.

What’s amazing technically is not an amazing code base.

What is an amazing code base isn’t an amazing product.

What is an amazing product isn’t an amazing business.

Literally every product I have ever seen that is profitable is not a good code base. But if it pays me to code, I love it.

If you live your life expecting high six figure income every year you’re risking catastrophe if you’re caught out in a down cycle. Keep expenses low and savings high.

Part of the high salary is to smooth out low or zero salary periods.

The easiest model to understand and prevent or detect burnout happens is when you continuously rely on willpower to work on things on which you have no hope of success.

Because hope is a story you tell yourself, you can blunt or escape burnout by reframing your situation into one that is entirely about internal rewards instead of external rewards which you cannot control.

For example, in a failing situation you are gaining experience and the chance to develop skills you can use later.

The difficult cases are when you have external dependencies that can’t be met like supporting a family.

When I was younger I always felt I could quit so I worked without fear and did amazing work. But when I felt I couldn’t quit, it was painful.

My entire life I have watched tech burn people out and I then got my turn after I had kids. The lesson is to remember that industry moves with the money printing. The high salaries are double reality and to keep expenses low so you can survive the financial cycle.

The expression is a social game to respond to death. It’s not real. We are not happy how he died. We are happy the way he lived. The game is to invert and mix the two. It’s a form of cognitive dissonance to manage our emotional response so it is positive not negative. You may not feel it yourself which is fine but appreciate that’s what others are doing. It’s not meant to be logical. It’s meant to be illogical.

It could be that Google has a professional CEO. In many of those cases, the customer is the board of directors. AI is driving share prices.

LLMs don’t excel at love poetry. However love poetry doesn’t have to be excellent to be effective.

I think of AI as a super keen arrogant intern. With GitHub copilot, an intern that constantly interrupts me.

When that works for me, I am probably weak on the subject material myself. eg writing quirky love poems to my wife in different styles.

For research tasks, because the AI is not deeply self-reflective, it can output inconsistent and incoherent results. What it does is present text that only *looks* as if it confidently knows what it is talking about.

For domains where high rational quality doesn’t matter like love poetry, it is amazing. For other domains, be wary. If you can’t tell the difference between what is actually good and what merely looks good superficially you will be in trouble.

That’s an interesting idea. What do you mean?

I understand the prompts as a service is short term… but what is a long term product you see?

There’s no reason. I just wanted to be cool and use neovim! Lol :)

True. My friend is selling electrified retromodded Porsches. I want one but I am poor.

However I can salve my ego spending a day flipping through neovim colour themes.

That’s fair. I can’t use vs code on the server was my logic. But also it was a hacking challenge.

I used nvchad and I am configuring it from there. Here’s my fork.

https://github.com/sunir/NvChad

Overall I still think I am faster in sublime text. I get stuck in the different modes. I find shift select and grep to be pretty frustrating.

However I will muscle through this. Every challenge is another set of vim stuff to learn. I have faith I will love it later.

I switched to Neovim from Sublime Text after trying copilot in Sublime, feeling sad, and then watching The Primeagen and his glorious mustache for too long.

Ostensibly I wanted to be able to code on the production server like a miscreant with the same tools as my laptop.

However I just wanted to regain command of my dev environment after years not coding.

I also reorganized the furniture in my office and got weirder lighting to make it hacker friendly. I bought a new desk to solder electronics.

Most people know me as a partnerships marketer or product manager but I am a compsci at heart. This made me happy.

Instead of being incredulous, assume good faith the other person is speaking with best intent. Try to understand first.

There are no free payment transfer mechanisms. Whatever is free is a loss leader.

If you don’t know anything about payments, start by reading stripe.com’s documentation to understand how complex this is.

Pymnts.com is the industry blog if you want to go deep.

Generally speaking with Ottawa these days, the default expectation to every time a promise is made that sounds good for the media is that nothing will be done after the announcement.

I feel that is a cynical take and I defend against cynicism; however I think the government is more cynical that announcements are the only thing that matters. So it’s a question of whose cynicism should I defend against?

The principle still matters. Learning how underlying storage works helps optimize data layout, write and read for speed.

It’s just that storage has changed. Caches may be important now for instance

Too much unproductive money was dumped into the industry during the pandemic. This was a monetary policy to keep the world employed and paying bills; tech and gaming and media had a rally because of the remote, online world we were forced into—many people believed society would change forever to be more remote and online.

Because this money printing and investing didn’t produce sustainable results, the businesses can’t afford to carry the payroll they hired with the invested capital.

So the bubble popped and layoffs are one of many negative consequences from having the distortion of the pandemic. Another consequence for instance is that people don’t necessarily know if they were good or bad at their jobs during this time because cash flow was disconnected from productive value; and I think that is a shame because it will mess up a lot of people for years.

ps. The tech rally had actually begun in 2016. It got supercharged in 2020. It’s unclear where things will settle.

You can start out by presuming they have reasons to choose a difficult path. Probably it’s anxiety inducing so they aren’t forthright about it but it’s not necessarily the wrong decision from their point of view. If you have the chance you could ask them what they perceive the financial picture looks like going forward.

To generalize without knowing your situation, right now people feel that no financing is coming for a long time if ever again unless you are a top quality company. So getting disciplined on cash flow positivity may be their focus for good reasons.