nasir

There's a fine line between being/not being that person. Self deprecating humor is usually works best. Also, when "everyone else is silently annoyed" that means the person can also know that. So there's a feedback loop providing signal and if someone doesn't realise that then they become annoying.

haha the dudes are of course the best. when I say "I can talk to anyone" it doesn't mean "Everyone will talk back to me". Which is fine and I don't care. For what its worth I'm glad not to have to talk to boring people.

What I want is to have a laugh or an interesting intellectual conversation.

Even though typing them out may make them stupid but here are a few examples thinking out loud. Remember the body language is quite important and as you do more you start feeling more comfortable in your skin.

- Waiting for an elevator that never comes with two strangers. What I may say: I guess we'd be camping here tonight. Do you have your tent with you?

- Embarrassing moment: I hit my head lightly to something in front of 5 people: Act funny saying Oh can someone call an ambulance.

- Someone dropping yogurt from their spoon on their shirt and locking eye to eye with me realising I've been watching the moment: I would have an empathetic look and then act with an imaginary spoon picking from my own shirt and eating it.

Basically the kind of mild jokes/acts you would do and say to close people would work on strangers as well

Around 15 years ago I took on the challenge to start a conversation with random people to break through this barrier and train this muscle. What I started with was to chit chat with those I had already established an interaction. For example at the Starbucks I would say something to barista. Those interactions were short but broke the ice.

Later I went for random people in the street and that was quite awkward. There was simply not much I could work with (what I thought at the time).

This turned out to become a low stake effort to improve my social anxiety. Helped me build humour around it and eventually become comfortable

Fast forward to today, I can literally talk about anything to anyone. The main pattern I follow is to break the pattern and make a joke, be sarcastic respectfully or give a compliment. No permission just something they don't expect. Almost works all the time. It helps with confidence and also makes you realise its all in your head.

And it is fun indeed

Mens sana in corpore sano

Taking the antidepressant has become the dominant go to approach as it is more commercially profitable for the seller and easier to commit to for the taker.

where's opus/sonnet! very curious on that!

I'm building a tool for managing Google Ads campaing called Rudys.AI:

- Search campaigns: - automatically crawl website, find the offerings and generate new campaigns - Provide qualitative recommendations such as, relevant terms to include/exclude, e.g. including typos as keywords, improvements on landing page

- Shopping campaign: - Smart labeling of all products to allocate the budget among Top performers, Rising and Ghost products to avoid draining the budget. E.g. instead of a campaign with 10k products with one budget , turn it into 5 campaigns with different budgets doubling down on what works.

Feel free to reach out for a demo

Building Rudys.AI to simplify Google Ads for non-specialists and reduce work for specialists.

Currently working on https://rudys.ai to publish and optimise Google Ads campaigns on autopilot.

The idea is to be able to publish campaigns globally in any location/language and also get qualitative recommendations on what to improve. For example, if people have typos in their search terms, Rudy recommends to add it as a keyword so it can maximise the conversion.

Learned quite a bit and seems like a basic but necessary thing to know about!

press esc twice, you get a list of previous checkpoints you can revert back to.

Using Fedora 3 as my first linux distro makes seeing this nostalgic.

I already run multiple apps on a single server with Kamal and it works fairly good. Sometimes there are some issues especially if you use similar resources across system such as redis. But overall it is stable and works good.

The chronological order is quite subobtimal. You still want to discover new content/ideas/conversation. It's easy to miss out on a lot of what your followers have been tweeting if you don't check it 24/7. On the other hand, when you sit for an hour to check twitter, the following tab will dry-out immediately and you can't discover anything new unless you go to the For You tab which is basically a mix of stuff.

Indeed if you're active, and like stuff, the algo improves and shows you lots of relevant content. But mostly from the prominent accounts and only viral ones. So the point of the post stands.

Living the same life. 10 years speaking English with a non English speaking partner and I can say I now fully think in English and in many aspects my English is better than my native language.

Could never put words on it but things don’t feel so real either.

The funniest part is now I speak my native language with my small kids and have fully relapsed to my original home town accent. I had abolished that accent years ago to avoid getting bullied at high school after relocating to a big city.

My kids are actively hearing and speaking 4 languages and one is with a thick accent!

Summary: "In a hierarchy, every employee tends to rise to his level of incompetence."

So I guess what I saw is to use a big model to do annotation and develop a specialised smaller model.

I'd be very curious to see how it performs especially on inputs that's blocked by other models. Seems like Grok will differentiate itself from other OS models from a cencorship and alignment perspective.

Django: function based views.

UI: Django templates, htmx, tailwind, vanilla js or hypersceript, no build, no node, no npm. Avoiding writing js as much as possible.

Background workers: RQ with 3 priority queues

Deployment: Kamal one command deployment with underlying docker

Hosting: Linode server with backups

Storage: Mounted volume on the machine (maybe s3 eventually).

Benefits:

* High speed of iteration

* No boilerplate mess

* No build therefore no dealing with dependencies and npm rabbithole

* Fast deployment with Kamal (a bit of learning curve but once up just works)

* Much cheaper compared to PaaS. At least 50%.

* Multiple websites on the same instance

Of course I know django quite well so the goal is to remove as much friction as possible.

It may be unpopular but speaking from experience, I think using boilerplates can lead to slower rate of iteration.

A lot of these boilerplate project make strong opinionated decisions which can be in conflict with how you build and structure your project. Meaning whenever you need to make modifications, you'd need to fully go through how it is done and could be counter intuitive to modify.

Also, for starting a SaaS, the initial version is probably better to be as simple as possible rather than a bunch of advanced features. Which means you'd be starting to slash features from the boilerplate at some point.

Paying attention to the micro interactions of a person (or CEO) with others especially those who have no business interest with them can give a good view of their morality. These can be simple things like a conversation they have with random strangers and their judgement on that.

I remember installing a Juniper Intrusion Detection System in a server rack on a telecom company. Was quite impressed when I saw in the logs such attacks were discovered and blocked. This was 15 years ago.