figassis

My process: start ideating and get the AI to poke holes in your reasoning, your vision, scalability, etc. do this for a few days while taking breaks. This is all contained in one Md file with mermaid diagrams and sections.

Then use ideation to architect, dive into details and tell the AI exactly what your choices are, how certain methods should be called, how logging and observability should be setup, what language to use, type checking, coding style (configure ruthless linting and formatting before you write a single line of code), what testing methodology, framework, unit, integration, e2e. Database, changes you will handle migrations, as much as possible so the AI is as confined as possible to how you would do it.

Then, create a plan file, have it manage it like a task list, and implement in parts, before starting it needs to present you a plan, in it you will notice it will make mistakes, misunderstand some things that you may me didn’t clarify before, or it will just forget. You add to AGENTS.md or whatever, make changes to the ai’s plan, tell it to update the plan.md and when satisfied, proceed.

After done, review the code. You will notice there is always something to fix. Hardcoded variables, a sql migration with seed data that should actually not be a migration, just generally crazy stuff.

The worst is that the AI is always very loose on requirements. You will notice all its fields are nullable, records have little to no validation, you report an error when testing and it tried to solve it with an brittle async solution, like LISTEN/NOTIFY or a callback instead of doing the architecturally correct solution. Things that at scale are hell to debug, especially if you did not write the code.

If you do this and iterate you will gradually end up with a solid harness and you will need to review less.

Then port it to other projects.

Hadoop is open source: https://hadoop.apache.org But perhaps if one learns Google they could have found it? It's a tough chicken and egg problem.

I think of it not as "last step in thinking", but as "first contact with reality". Your mind is amazing and lying to you, filling in gaps and telling you everything is ok. The moment you try to export what's in your mind, math stops mathing. So writing is an important exercise.

This is fine. But companies seem to not have a control lever for employee wellbeing. If humanity works to solve problems, don’t you think overwork is also a problem that needs to be addressed?

Where are the lessons?

“The rationale for the change, according to Rodriguez, is that interaction data makes company AI models perform better. Adding interaction data from Microsoft employees has led to meaningful improvements, he claims, such as an increased acceptance rate for AI model suggestions.” Things the problem with promises. Of course using the data makes your model better, and everyone new this and that you’d be tempted to use it, and that is why you felt motivated to promise not to use it in order to gain adoption. So rug pulling by saying it helps improve your model is meaningless. You can say you lied. Lying is a real word, it should be used when it applies.

This just means we're in a simulated universe. He's respawned elsewhere.

I think apps that want assurance of your identity should pay for your kya. They want valuable people after all, and this should go into their CAC. Users still pay nothing, the identity service does not care about their info, after verification, it drops any details, like uploaded documents, whatever, keeps a certificate.

The cost for this service is likely keeping up with ID systems for multiple countries, infra and support.

Potentially, if this is made into a protocol, it can be decentralized kind of like the SSL system, so each country manages it's own rules.

I am less concerned here. If you plug in AI into your identity, I guess your identity is revoked. I see the problem though, that once a service notices you're an AI, there is no way to block you because we don't really know who you are, only that you're human.

So we need a mechanism that makes this identity verifiable, maybe you get a unique identifier from the identity service, so you can block the account. There is no mechanism to report you to say, the identity service (this is a bot), so you manage your own block list.

The risk here is fingerprinting, your id can be cross referenced across apps. Maybe here is where you implement a zk proof that you're who you say you are.

The problem seems to be identify, a real problem, and looks like it will only get worse. Would creating a zero knowledge digital identity service (maybe centralized, maybe decentralized idk), where you prove you're human via your government id, passport, driver license, whatever, and the service can then attest you're a real person? So if I'm Digg, I would ask for some form of OAuth, the system would simply verify that you are in fact a human, and you would go on to create your account, forever verified. This way the identity service only does identity, it does not keep a record of where you are attesting, no logs, nothing, just your identity and basically saying yes/no, no sharing of ids or any other data.

So people would go through one hurdle in life, to get this id, and reuse it for every service.

Is this a worthwhile idea? I know many have tried, so help me poke holes in it.

Can we just use something like AWS instance roles where the key isn’t even known to the agent? The sandbox is authenticated, the agent running there sends requests, the sandbox middlemans the request to perplexity, which authenticates the sandbox.

I see how this can boost productivity...for those that today already produce value voluntarily. These will move one level higher. The rest with 100x the amount of performative work. Everyone will be busier created presentations and charts that no one needs and no one will read. Managers will ask for new presentations and reports every sync, and hours will be spent discussing things that don't actually matter.

I think with the amount of AI content, it will all become cheap, there will be movie releases every day, people will get content fatigue, it will all merge into the same thing, nothing will be inspiring anymore. People will start looking for physical experiences. Movies that transcend the screen, think 4D experiences, but all the way, putting you in the center of the action, maybe you're the protagonist, etc.

I'd like to think I am pretty security conscious, but I still don't get the obsession with magic links (and passkeys). This is the one thing where I think I disagree with most of the industry. I thought forgetting passwords was a solved problem. I thought 2fa is much faster than searching for the last email for X provider the maybe takes 1 minute to arrive, requires retries and high tend up in spam? Some one please help me get on board.

I would create a library, make some logic more generic, create a generic table (task id, taskType, workerId, etc), store task metadata as jsonb so it can be pulled and marshalled into typed data by users.

Import it into your projects.

Make the library work standalone. But also build a task manager service that people can deploy if they want to run it outside their code.

Then offer a hosted solution that does the webhooks.

I’m sure someone will want to pay for it.

Use Go, it has built in go routines and likely libraries that let you implement your own workers.

If you’re running a single instance, you don’t even need any synchronization. If you’re running multiple instances of your app, try implementing locking (this actually works in any language, not just go. Go jsut helps with the multiple long running workers part. With other languages, just run multiple instances.

Process:

1. Each worker can startup with their own id, it can be a random uuid.

2. When you need to create a task, add it to the tasks table, do nothing else and exit.

3. Each worker running on some loop or cron, would set a lock on a subset of the tasks. Like:

update tasks set workerId = myUUID, lockUntil = now() +10minutes where (workerId is null or lockUntil < now()) and completed = false

Or you can do a select for update or w/e helps you keep other workers from setting their ids at the same time.

4. When this is done, pull all tasks assigned to your worker, execute, then clear the lock, and set to completed.

5. If your worker crashes, another will be able to pick it up after the lock expires.

No redis, no additional libraries, still distributed

1. Development is now so cheap! Pay us the tokens!

2. Development now creates unreliable code at incredible speed! Pay is more tokens!

3. You can’t reliably verify any code before prod, pay us all the tokens so we can do it for you, possibly also unreliable!

It used to be that when you pay for something you expect some level of quality, not the appearance of it.

The mental acrobatics today are simply because engineers have become expensive. But engineers didn’t make themselves expensive, tech companies created this culture in order to monopolize talent, and this caused inflation, which leaked into the cost of living for these centers of gravity, which resulted in even more wage inflation, perks, etc.

They can’t solve this problem now, are blaming it on engineers, the government, w/e.

The solution seems to be, clone cheap engineers. They will improve later. I don’t have a problem with this, a valid idea is a valid idea.

However, extrapolating to the logical conclusion, It will be quite interesting when companies realize the trap. Vendors will jack up prices so much that companies will scramble for engineers, which by then will have moved on to other careers because jobs dried out.

At this time, LLM vendors will be offering refactoring solutions. For $1M, we will rebuild all your apps behavior from scratch.

Then they will tell companies to give up code altogether. They will sell AI as a service where you give it an idea, the ai builds it internally, you don’t get to see it bc it’s not your IP anymore. You only get to validate behavior.

Then, the next step will be for the AI companies to cut the middleman (the client companies) and go directly to their customer base because they understand what they need better than anyone, that an app is unnecessary. No UI is needed, you talk to your AI companion and things magically get done.

No one will know what a spreadsheet is anymore, some will forget how to use a keyboard, and at this point some laws will have been sleepwalked into maintaining this status quo in perpetuity.

It’s very easy to convince an addict to buy more drugs.

OpenAI new this would happen, so they hired the creator first.

That is fine, they're not required to support unexpected use cases. Not the same as forbidding you from using it as you see fit. It's simple really, they can do what they want with their resources, you can do what you want with yours, especially those you paid for.

“Why is the Patek Philippe Nautilus so expensive?”

If you have to explain why your product is expensive, maybe it shouldn’t be.

He is misinformed. Employment laws aren’t new, they’re the laws these countries adopted when they became independent because that is all the law they knew, since when they were colonies, these were the laws they lived under, and similarly to Europe, they didn’t know better. Also, neither does the US. I get the downside of these laws, but the jury is still out on whether lax employment protections are the solution.

I have just done a performance audit on some endpoints of a very fast moving team. Man, it is disheartening. A simple GET request to fetch one record has loops in the controller with nested database and external api calls.

And this is in a well architected and documented repo, where directness is prioritized over magic, with little to no useless abstractions, on a typed codebase, with development guidelines that address these sorts of specific issues.

Ironically, if you ask it to figure out why this endpoint is slow, it will answer correctly.

So most NATO people will have access to this level, so it's where most devices will be sold.

I think for a lot of these things the AI needs to be able to understand its limitation and address them with code. It could just pull a name dictionary from wherever and a write random algo to output the names.

They are forcing bc the goal is to reduce the cost center that SWE is, and are aware engineers will fight it because we don’t think it’s there yet, but they don’t want to wait for it to get there, they want to feed us into it so we can all train it to get there.

Your chosen tools didn’t have business impact, you did. LLMs have (or are thought to have) direct business impact. So you’re no longer in charge of deciding whether you use it.

You don't get hired for any of that. OpenAI did not hire him because he went viral. Virality brought something interesting to OpenAI's attention. And they thought they could use this product idea/vision/execution, GTM strategy, whatever, and because it didn't seem like OAI had anyone on their team capable of this, they hired him.

Simple as that. Don't feel jealous, trying to replicate won't work, he did not know he'd be hired, he built something that he found interesting, and then realized it would be interesting to a lot more people.

The way to reach success is to either be strategically consistent in a way that maximizes luck surface area but does not depend on it, or to be unexpectedly lucky. The latter is gambling, People win the lottery regularly, does not mean you should make that your mission.

Be comfortable with not being the one to hit gold. And yes, it's ok to be jealous. Take a moment and then go back and enjoy the rest of your life.

Finally, there are a lot of companies that would likely hire you, hoping to hit gold. But you are likely filtering them out because they're not tech/large/startupy enough for you. These companies are wondering what they need to attract talent like you.

Great list, missing zed on IDEs

It's not about being recognized, it's about when you are asked to be patted down, having the courage to lie "I can't raise my arms over my head", knowing the risk of being caught is at worst not making this flight. For a foreigner it might be getting banned permanently from the country. Same concept as self censorship. You do what you're told and then you go enjoy your vacation.

It is mostly security, but not to residents of the country. Those can enforce their rights. In my country, I can argue with airport security, and win. Foreigners can’t, so they follow whatever rules. A few times when landing in the US, security was extremely rude, I think just looking for an excuse (things like throwing your laptop a few feet away, while staring at you, etc). You take it bc you’re not home, and the cost of ruining your vacation is not worth it.

What I’m trying to say is that , while a lot of it is theater, TSA may be more effective security against foreigners but you as a resident don’t notice because you can opt out. Try going to the UK and telling them you can’t raise your arms while being a US citizen.

What part of faster is better means engineering to you? Non engineers will prefer you get there faster, but however you get there, better is better.