Software engineer with impostor syndrome

Remote
Joined May 2021
my birthday is on october 14. so naturally, i built a website where brands can buy me a MacBook. in return: → your logo on my MacBook or accessories → appearances in my videos → replies under my posts for up to 12 months there are only 20 sticker spots and you can bid on them. i don’t have the MacBook yet. that’s what the bids are paying for lol auction ends on my birthday 🎂 buymeagift.lol
83
28
11
127
28,979
LLM are slot machines! You continuously try luck with different prompts to AI, and sometimes you win by receiving the correct response.
LLMs usam estratégias de caça níquel. - Sorte de principiante = Te entrega muito no começo e depois vai caindo. - Recompensa variável = É o mesmo princípio dos feeds que te mantém viciado, pq a expectativa te gera mais dopamina que a própria vitória. - Ganhos decrescentes = Sempre te entrega muito no modelo novo, vai até surgir outro modelo novo. - Efeito quase acerto (near miss) = acerta 98% e deixa propositalmente algo errado pra fora, isso te faz gastar mais e ainda te gera efeito IKEA porque tem a sensação que construiu. - Efeito autonomia = Tanto uma máquina caça níquel quanto uma LLM poderia rodar por horas depois de uma tarefa recebida, mas por padrão optam em te oferecer alguma opção para te fazer achar que tá no controle - Efeito velocidade = Rodadas rápidas encurtam ciclos e possibilitam mais jogadas, a LLM faz isso naturalmente em alguns momentos. - Dissociação do dinheiro = Toda vez que você distancia as pessoas do dinheiro (tanto físico, quanto mental) as pessoas gastam mais, por isso essas contagens malucas de tokens que arbitrariamente mudam de valores te deixam confuso. LLMs são incrivelmente úteis e talvez a maior mudança tecnológica da nossa geração, mas oq a OpenAI e Anthropic fazem é usar estratégia indutora de vício e perda de referencial já conhecida há décadas.
3
2
1
37
2,247
A friend did a thing, and it's an idea I can totally relate to, so take a look.
🔍 Search for a free tool online. Any of them. Pick one. ➡️ Close the cookie banner. ➡️ Close the newsletter popup. ➡️ Close the "create a free account to continue" popup. ➡️ Close the ad. ➡️ You wanted to resize an image. I got tired of it, so I built an app that consists of tools that I frequently use. Introducing ⊞ zeal.tools — no popups, no sign-up, no ads 🎉⚡
4
1
14
651
Be honest
2
2
5
171
Roman retweeted
Replying to @Swayson @romxdev
In einer grossen Studie wurde untersucht, wie sich die Leistung von über 26'000 Schülern in China während 30 Monaten veränderte, wenn sie anfingen, KI-Chatbots zu nutzen. Ihre Hausaufgaben wurden rund 20% besser. Sie benötigten für die Hausaufgaben rund 20% weniger Zeit. Das ist super. Aber: Bei Prüfungen (wo KI verboten ist) wurden sie rund 20% *schlechter*. Das ist eine massive Verschlechterung. KI kann Denkkompetenz aufbauen, wenn sie als eine Art Tutor eingesetzt wird. Dann spricht man von kognitivem Scaffolding. Die Realität ist aber, dass die Strategie des kognitiven Offloading der Weg des geringsten Widerstands ist: Denkarbeit an Chatbots auszulagern, ist instrumentell gesehen rational. Ein Fehlanreiz. Diese Entwicklung ruiniert Bildung. Und sie ist ein systemisches Risiko: Was passiert, wenn eine ganze Generation noch weniger als frühere Generationen lernt, eigenständig zu denken?
1
1
2
137
You might believe you should spend less time thinking about code because of AI. I strongly disagree! We’re watching this play out live where tons of AI generated code becomes a liability. At the end of the day, an engineer needs to be responsible / on call for code that gets shipped to production. If you don’t understand the system you’re trying to debug, you’re probably going to have a bad time. Yes, AI can help with all of this, if you set up the proper systems. You can have agents triage prod logs, look at errors, etc. You can speed up parts of the investigation, but an engineer needs to make the call. There might be serious customer or financial implications from that change. I expect the trend continue for trimming dependencies, vendoring code so you can modify it directly, preferring simpler systems with fewer abstractions, and spending waaaay more time thinking about system design and code maintenance. I’ve said this before, but it’s a great time to get familiar with CS fundamentals and some of the history behind what great software looks like. Many parts will be different in the coming years as AI progresses, but also a lot more than people realize will stay the same.
262
519
164
4,150
610,109
Today we're also rolling out a tool to snooze topics on your For You tab—if you ever want to crank up or turn down the slop. Rolling out now on iOS and Web for Premium subscribers.
Ladies and gentlemen, today we're launching one of our biggest changes to 𝕏 Introducing Custom Timelines This feature allows you to pin a specific topic to your home tab. With support for over 75 topics, you can dive deep into your favorite niche on X. It's powered by Grok's understanding of every post with the algorithm's personalization—meaning every timeline is made just for you. And it works even better when it's a topic you already engage with. This was a huge undertaking across many months, so we're excited for you take it for a spin. We're giving early access to Premium subscribers on iOS (and Android coming very soon).
1,488
634
383
7,091
2,287,498
If this is really generated with the brand new @OpenAI image model, then we're in for some good times. This might be the top model for marketing use at least until Google drops Nano Banana Pro, which is most likely to happen at Google I/O at the end of May. 12pm PT so it should be 9pm in Central Europe and 12:30am in India.
This is not a screenshot.
8
1
11
440
cunning
85
78
14
7,541
432,944
This is currently most important for us, casual users, from all the @AnthropicAI announcement today. • Mythos Preview will only be available for critical software (like AWS stuff) and key open-source projects, to help with security reviews and bug fixes. • New Opus incoming, either 4.7 or 5.0, we'll see. • Rumor guys point at middle of the month, but they also predict DeepSeek everyday.
Replying to @AnthropicAI
We do not plan to make Mythos Preview generally available. Our goal is to deploy Mythos-class models safely at scale, but first we need safeguards that reliably block their most dangerous outputs. We’ll begin testing those safeguards with an upcoming Claude Opus model.
2
1
12
530
Replying to @romxdev
Vibe coding should be renamed to Fast Code, which is the equivalent of Fast Food and has similar compounding side effects
1
1
1
465
Replying to @Trader19731
Yeah, the world is changing.. not always in a good way.
1
11
1,159
Roman retweeted
We just released Gemma 4 — our most intelligent open models to date. Built from the same world-class research as Gemini 3, Gemma 4 brings breakthrough intelligence directly to your own hardware for advanced reasoning and agentic workflows. Released under a commercially permissive Apache 2.0 license so anyone can build powerful AI tools. 🧵↓
714
2,954
1,106
20,074
7,761,469
Plan for April Less random spam More quality content More helping out others More time in my friends' timelines More focus on making real connections Being nicer to people and bots
7
3
25
2,118
The difference is that @AWS is serverless, while @kfc is more like serviceless.
3
1
15
384
Roman retweeted
Brave
@braveMar 24
Brave just registered a .agent domain! We support the effort to have the .agent top-level domain managed by a community, instead of being owned by one company. Join the community and pre-register your domain here: agentcommunity.org
129
249
121
2,998
544,110
Can we all tweet like a normal human being? can we go back to the days when jack Dorsey still own the 𝕏 street? Can we please tweet our thoughts for fun? I’m tired of seeing tweet for impressions…
2
1
3
61
These are my super active followers @SdqAlvi @Bellorris @ViralHaseena @MonetXpert @siggres1 @Queen_Phatsi @prince_of_web3 @eth_2017 @Zarka_khanx3 @engniiokai @tblinz @VibeWithJuls @KarinaVibez @Xusa00 @ML1_GR9 @web3queen__ @romxdev @P0PSKI @Truelvo @Bannor04
188
52
1
210
5,596
I treat people with respect regardless their age….
1
1
1
55