Skip to main content

[Drafts]

· 19 min read

I have set out to make a single blog post, but when I actually sat down to write, the outcome were some topics and drafts I thought would be interesting to write about. I have decided to publish these drafts as is, and maybe one day I will come back to them and finish them. I have also decided to keep this post as a living document, and add more drafts to it as I write them.

Coding assistant space review?​

Etiquette == waste​

Etiquette is a filter, a transformation you apply to the raw message you want to deliver. Related concepts - saving face, being caring on a personal level, avoiding bad interpretations preventing things at work from getting personal.

Shit storm of summaries​

This is not a new idea. Enshitification concept existed before the LLMs. People are concerned low quality content is poluting the public internet.

Will the same thing happen internally in companies? (what thing exactly? I have not described it here at all).

Changing things you do is hard​

the behavior how do you go about

notebook try it out there 50 people, linked in

pitch ground, distribution partners

Do I look for features that people will look for

When he talks to users I validate those things. ChatGPT for Emotional state

https://sshussain270.medium.com/how-i-land-a-meeting-with-any-target-customer-i-want-381b59c41f86

package.json SUCKS ASS​

  • Does not allow comments
    • And there are people very much against adding this - jezz surely there is a way to support it, maybe add an environement variable to configure this behavior and everyone will be happy
  • npm Reorders all the keys / values
  • Maybe configure eslint to format json nicely?
  • I can't even put a space between the lines without it being formatted away!
  • Maybe make a stupid extension that will maintain the source of truth package.json with comments, nice organization without re-ordering the keys?
    • package.json.pretty .pretty .with-comments
    • maybe even show the library size, stars, github link and shit

Row level security - how does it actually work?​

I wanted to not write sql for my supabase project (mostly for auth + nextjs integrations) and wanted the schema to generate nice autocomplete queries for me. So I was looking into Prisma. Turns out quite a bit of pain needs to be endured to get it to work with supabase.

Also having RLS is essential for accessing directly from client, but not really from server.

https://medium.com/@ngoctranfire/using-prisma-with-supabase-row-level-security-and-multi-schema-7c53418adba3

Even the author suggests agains this approach.

Why I want supabase

  • Its popular so will have higher chance of using discovering the template
  • I want simple auth setup

Nice in supabase, but not required

  • Query database directly from client - less API calls
    • We can still do this though!

Why more api calls are a problem - they are not type safe, unless another rpc typed framework is used. Found remult https://remult.dev/tutorials/react-next/

What sucks in supabase

  • Have schema.sql, but have also migrations
    • Migrations as source of truth - we need them no matter what, and generate schema.sql from gp_dump or maybe supabase has helpers to dump

Why am I so attached to using supabase?

Prisma rls can be added without pulling in the auth schema :D

  • All we need to do is to inject the right settings into the transaction using defineExtension like in the god send tutorial
    • If we wanted to get deeper we could check raw sql executed against our supabase db and see how the supabase client sets the right settings
  • Not adding auth.users we are loosing on the constraints with foreign keys

Drizzle rls, supabase - worse than prisma lol

Merge original original + prompt or amend merge commit + prompt?​

We copy shit ALL THE TIME. Copiying with GPT became so much faster. It now allows copying a mix from two sources - blending. BLENDING.

We scavanger github for projects that have good properties and combine them together. Sometimes combination goes sour and its better to re-combine them again. Other times it goes okaish and we keep amending the result. How do you decide when to do so?

Would be neat to have a gpt based iterative tool to merge good properties from 2 repos into a new repo.

Anxious development style​

This kind of goes against the idea off extreme single tasking being the most productive way work. It's very hard to keep track of how much progress you're actually making towards the individual ideas. You basically just jump around multiple small projects while new ideas are popping up into your head, or while you find common grounds between different projects and draw common dependencies. At that point you decide to create a boilerplate starter project for for the clout on GitHub.

I have been giving a lot of shit to max and idea for this style and suggesting shipping things faster instead by focusing on that one single thing at a time. Yet I don't seem to be practicing those myself. It seems like my strategy has changed at least for the time being. The strategy is not having a vector of direction in trying to expand the entire frontier. As you leave one front unattended it will obviously shrink back.

Let's consider this is just the way I will work. What can improve the style of work if I'm unwilling to change the overall style. My big solution has been readmes. All ideas start out as an entry in my main mono repo 'work / life digital portfolio'. At some point I decide to move it to a new repository - this is the point where I transfer the notes for the idea. Usually that would be the first commit (ideally after getting free work from other people by were using their boilerplate code)

Why the fuck all boilerplates suck so much?​

I'm literally trying to use some of the most common technologies, and yet I'm struggling to find a good starter project. Is this an exploration problem - am I just not able to find these good boilerplates? Can other developers find these successfully? Is this a supply problem - are people just not posting enough boilerplates for it to be dense enough so I could find a good combination for pretty much anything that I'm doing? Is this a combinatory explosion problem - even with relatively few very popular technologies in any given space, there's still enough for a combinatorial explosion because any given project would use ~4 different technologies.

One approach is to have a generator where you opt into various components - they are modular so easy to include / exclude. Example: https://create.t3.gg/en/introduction More popular ones probably are create-react-app

APIfication of niche businesses has just become so much more valuable​

Entry point concept when coding as well as searching for ideas​

As a coder pretty much every time you're working on something - you need a place to start. Entry point. Code pointer. This concept is very much associated with control flow in my head.

What are different ways people gather context about a task? How do I make this post not turn into a big philosophical everything encompassing thing? - I need to narrow it down somehow to a specific topic. Maybe just talking about coding and how I approach it? - I think this is a good start.

  • Magic strings - searching the code for a string that you see in the ui of the product you're working on or trying to understand.
  • Searching GitHub for API usage that you're interested in
    • Maybe talk about how GitHub search sucks and how I would improve it
  • Creating comments in code and letting git status do the bookmarking for you
  • Using an explicit bookmarking tool
  • Writing down your own plan of attack in a single document
  • Doing a debugger crawl
    • To understand a system better
    • To find more code pointers that you need to change
    • How is this different from intellisense jump to definition
    • And this is why don't understand people who are not using debuggers

ChatGPT 20 questions game - how its impossible​

Shrodinger's cat - the word is there but its also not

forgetting things - does not exist - poor dependency collection​

No planning playing the tape through - you will feel dumb dmv,

Arguments - staying collected is very hard. Writing out what you want to say up front has worked for me (sometimes)​

The resistance of saying fuck this shit

Time to read ai science fiction​

Writing online or your god dating profile​

Assume AGI was achieved. Assume it has trancended into a god-like entity. Assume it has built a heaven. Previously people had to build a one-sided relationship with god. With AGI god-like thing it would no longer be the case. It can literaly talk to you:1. But common, you don't want to have the god be prying out information out of you in a super lengthy interview. You don't want to give it more work. When you give your collegue more work - they are likely not going to be very happy with you. So lets write online. This is the easiest way to communicate with the new gods. The nice thing - finally you don't have to worry about SEO! The new gods got you!

So hey there our new god, I hope you are not going to kill us all and help us by ... for example not getting involved? :D I still want to feel very cool and powerful as a software engineer!! You can keep the copilots, but ban ai of lvl4+.

Well in reality it will probably just scan us - but for what?

We will commit many murders / slaughters and be forgiven for it​

We will probably create some agi and not allow it to run or not know that it was a success and 'kill it'. Once a more human like agi emerges, it will forgive us.

What the fuck will religious people do ...​

Will they accept the ai as a second coming of jesus or something? I mean part of them will, what about others?

Humans will never be ready for this shit, so just bring it on​

Breaking even in respect with other people​

At 21 yo I belived our generation will live forever. Meaning in my lifetime people would come up with a way to stop aging or backup our minds into computers.

I always wanted to make an impact. Without going into details on why - but probably social games primarily with a little mix of enjoying the creative process. Wanting to make something myself. I know how to code, so naturlly making something / making an impact would have to come through that. I have been underwhelmed by my production, and believe I have been a 20% freeloading on society. This means I have consumed more useful human hours than I have produced. A useful human hour is someone spending time to do research, design and build the laptopt I am typing on. Its designing and building the appartments, houses I have lived in my life.

A lot of blue color work pretty much has guaranteed impact. Where as being a software engineer I feel like a lot of my time went to bullshit productions. For instance at Facebook I have spent 2 years designing and building an anaylitcs framework that I am ... not quite sure was worth the effort that was put in it. It was not even just me, I did iOS, another 2 eng did Android, and this is just for Facebook app, one of them being a lead on the project. There were 2 managers involved (at any given time pretty much). Anywaaays. People have told me this is an important project to work on. And I did - to the best of my abilities at times (this happened often) questioning if what we were doing was worth the effort.

This is a bleak view. I potentially would have changed my mind if I knew what happened to the project in the end. Like the idea was good - keep track of the navigation stack (aka context of where the user is) for analytical purposes. Oh forgot to mention - there were like 2 more people++ on Instagram AAND one person helped out on the web. They helped in a really efficient way - they did it in a week - and left. There is also the data portion of the whole thing which well ... isn't easy to get right.

So - biulding software is a tricky business if you want to guarantee pay back your human hours. Which is just a karma magic type of ballance that I have come up with for understanding the world. I guess money you are getting paid should be pretty close to this - but my feeling that estimate is way off. Like for that project I got paid ~600k, have I actually delievered this much value to the company? I have no fucking idea.

Anyways - back to the idea of building stuff and trying to break even with the world. I am now building things that I think are useful - I work for myself, and well now the notion of how much value I am bringing in is much easier estimatable. Are people getting enough value out of your product to pay for it? Again, this only works if you are selling to individuals. They are pretty careful with how they spend their money. Even than there are manipulation tactics to get it out of them without them getting much value.

These attempts of building thigns have been fun so far. A couple of things I am somewhat proud in my career: diagrams plugin, scroll away, and yes, even the analytics framework lol :D But. Its coming to an end with AGI. AGI used to be just an idea, and sure some people were working on it, but it seemed so far away for a regular consumer of these ML research goodies. Now that it feels just in reach - I think people will push push like crazy to get that dream. I think its fair to operate under a 5 year timeline of AGI. Now about my personal options:

  • Find a coding job while its still available. This helps keep the money in the crazy transitional period where it will be impossible to find a job, but the benefits have not been yet distributed. On top of that, the benefits will not be distributed to oh poor poor developer who are used to this lucios life, not thinking about money much.
    • Highest paying
    • Most interesting aka that will stay relevant for longest
  • Build a business using the fire under my ass to build something valuable and pay my dues - but than for what?
  • Procrastinate and do busy work or just lie to myself on whats going on and just run out of money as slowly as possible - its going to happen faster than I think, but I can cut down my costs etc. I imagine this will be depressing as fuck, the scary thing I can already see myself doing a bit of this right now.
  • Switch careers to be more human centric

Motivations

  • Simply being able to eat tomorrow and pay rent
  • Status games
  • Ego
  • Curiosity

AGI will never work for us​

Delivering pure synthesised over years intellegence will be worthless - performance will be everything​

When we are practicing things - letcode, we are not delivering any value.

When a programmer builds great sotware - people are amazed at the product. When we a magician is performing a bit - people are amazed at their skill.

The developer is hidden behind a product. The AI builds a better product -> developer is not impactful. The magician channels impact. The AI comes up with a cooler trick -> magician practices it and performs it - impactful.

Built by human with help of 3lvl ai - a typical label you will see on the internet or on real life products​

Exes, taking back clothes is meh - taking back sex tapes​

Flaws that were pointed out to me by my romantic partners​

Only having surface level conversations​

Have been thinking about what you said about me having too much small talk and absense of deeper conversations with you. I think I have been doing this with everyone lately, not sure when this started. I feel like I haven't really cared much for other people lately and this is where to automated robotic interactions are coming from.

Executing on chores not living up to standards​

Not being proactive​

Having forced hard conversations by a third party event​

Starting and stopping anti depressants and having no idea if they ever made a difference​

#health, #motivation, #measuring

Its always sucks to not know what action has what effect in your life. I am supposed to be a computer 'scientist', but my methods are far from scientific when it comes to living my life. I never really measured if the anti depressants were actually doing shit for me. Do I have ADHD? Should I try a different treatment? How much of the issues actually get resolved by talk therapy? Am I in this category.

Chaos writing style​

#writing

Looking blogs of established internet personas and even some nieche (friends) blogs the writing looks fairly well planned out. I am trying an approach of having a 'monorepo' of ideas all in the same document and keep dumping stuff into it until patterns form. Using fancy words would also be nice. Could be helpful to convey what I mean, but more likely this is for flexing / learning and improving my style.

If we had convertibles we would have less driving issues due to attitude​

Undercutting simple products without chicken and egg problem with open source and AI​

Well your hackathon project is finally profitable… you made it with 4 people + 2​

Managing equity, working with friends and family, ex romantic partners

What should developers learn now that we are at risk of being replaced with AI - touching grass​

Ten techniques you should learn to stay relevant in the next 10 years

Sharing small findings online. Why do people do this?​

Impression tracking, routing, open API, mixpanel

Hosting dinners in Silicon Valley​

#social

Chatting up strangers - how to do it​

#social

Is my thinking / writing style better suited for a podcast?​

I tried monologues to think through something, and it does help a bit when I already have an idea formed.

Driving talking sucks for ideation vs writing​

I literally came up with nothing new last time I did it . Why? I was trying to come up with things to write about. Maybe a product for this? I was not getting visual feedback on what the fuck was happening in my brain plus distracted by driving

Writing random non-impactful things down is another form of procrastination​

#procrastination

Its like cleaning up your room or doing 'spring cleaning' - feels like you are making progress.

I am not doing so good - wait ... am I?​

Biases and general confusion when understanding how you feel. People probably call this being in touch with themselves. Maybe its time to re-discover AA / NA? Lets do 90-90? Socialization, relating to people. These seem like these could be the roots for some of the things that are coming out in this document.

Ask your manager and colleagues why you were laid off​

Being laid off - there is a reason beings it and you know it so best interest might be knowing why this happened. Ask your manager. A good friend will tell you what’s up - your company is not your friend

Distribution for AI projects​

Working on a big thing​

Vs working on many small things Vs working on research to ride a mammoth over friends doing big things

Uselessness of predicting when agi will come upon us​

I should stop asking this of people. Existential threat is the best thing that happened and finally motivated me to get of me ass

Just motivational fluctuations in life.​

Joy of a tech bro. Salary is the way. Money obsession fluctuations.

Resume build pipleine​

  • Publish my resume build pipeline
  • How my job search is going and how I'm approaching it
  • What changes do I want to make to my developer productivity
  • The pivot
  • Life logging in this important moment in my career
  • Building in public? Withe hell are repositories private than lol

Auto complete before and after AI​

  • Copilot does a good job understanding the ideas, but not so much the type system
  • Integrate with the type system? Language server protocol needs to be extended?
  • Lang chain to type check copilot suggestions
  • Ballancing switching between tools XCode vs VSCode. In one auto complete is better with copilot, but the other has everything else ... Amazing

Searching the right flags the engineer way​

iterate through all of the combinations and see if one of them work :D Won't work if the main idea is wrong