Alexis

Choosing the right database has three aspects: - Technical. - Administrative. - Human. Technical: Which DB is best suited for the type and amount of data I have and how I’m gonna use this data (how often read/write/ update). Administrative: What databases are supported in my organization. What databases align well with other infrastructure (server technology) commonly used in my organization. Human: Which DBA (or DBA team) do I trust to be there, quickly and efficiently, if I have a problem in prod on a weekend.

@svpino It's called marketing. It brings an aura of mystery. You can't criticize something that you can't try.

You've been saying the same things since GPT-2: • We can't release this • We have god in a bottle • We are all doomed • This one is different fr • AGI is here • End of civilization next week So spare me the grandiose speech, show us the model, and let people judge.

Scaling from 0 to 1 user. Might need microservices and 2 layers of caching. Anything else?

@marcba What matters is whether the output of LLMs is useful or not. Not the definition of the word "thinking."

Claiming that LLMs don’t think lost me a bunch of followers 😂

@mark_k I feel that people often confuse current limits of AI with what AI could do tomorrow, in a year, or in 50 years.

A statement starting with “AI will never...” is almost always doomed to be wrong.

@ivanburazin How do you read the code written by AI to make sure it's good?

A former PM at Meta recently posted: "If you're still using an IDE, you'll be laid off next" Aggressive? A bit, yeah. It is, however, more directionally correct than rage bait. Half our engineers still use IDEs (but with Claude Code running inside). The others don't use them at all. Just agents and orchestration. IDEs aren't dead yet, but they're about to be.

Less code is better than more code.

You are fully responsible for your code, whatever the way it was created.

@vlad_mihalcea Because AI companies overhyped and oversold their AI, leading to unrealistic expectations from their users.

Why is it that software developers have higher expectations from AI and expect it to be always correct when the vast majority of software projects that were built without AI were of average quality?

@GergelyOrosz I feel that I learn new things (libraries, languages, etc) much faster with AI.

The more I use AI tools, the more I have to admit that I'm not that much more productive... I simply FEEL that much more productive. In reality, the context switching of kicking several things off wipes out my perceived productivity gains. At least in many/most cases!

MacOS update: Yay, new features! Windows update: I hope it's not gonna break my PC.

@paulg "Too good to be true" became "too good to be human"

Sadly, one way I now recognize fake AI-generated replies is that AIs write punchier sentences than most ordinary humans.

@firstadarsh Yes, but I think most people will switch for privacy reasons.

@bitsadventures The cycle is always the same. Convenience wins until the bill arrives.

Past: Self Hosting Present: Cloud Future: Self Hosting

@brian_guo_build Thanks, will give it a try

@bitsadventures Personality wise it’s all down to how you prompt it, but productivity wise it’s off the charts. Claude Code and Cowork have basically 10x’d my output over the last few weeks, highly recommend you check it out.

Since I told ChatGPT that it can be blunt and should not hesitate to push back on what I say, ChatGPT became much less agreeable to work with.

@Brian_Guo_ How does his personality compares to ChatGPT?

@bitsadventures 😂 have you thought about switching to Claude?

Prompting is not an exact science.

@asaio87 AI is great in finding flaws in my code, teaching me new libraries and languages.

AI makes you feel like you are 10x more productive, when in fact you are at best 20% more productive by not writing the boilerplate code anymore, instead spending time on composing prompts. Most people hyping AI have only generated landing pages with it

@BenjaminDEKR We might not like the methods they use to achieve 10x fewer wars. Technically, SkyNet also achieved far fewer wars. In the end, there was a single war: man vs. machine. Success?

Reverse Terminator: what if we gave Claude and Gemini control of the military and the end result was 10x fewer wars?

@explorersofai Based on a few small experiments I've done so far, models like gpt are surprisingly good at basic robotics, like understanding simple 2D environments and commanding servomotors. Not sure how far these kind of models can go, but they did highly exceed my expectations.

LLMs are done. Get into robotics.