David Reis Profile
David Reis

@dave_reis

Followers
267
Following
583
Media
62
Statuses
3K

Software Engineer Founder at Vom Formerly CTO at Hubla, and Sr Staff Eng at Google Opinions are my own Read my writing at https://t.co/hRSPnJAcYm

Belo Horizonte, Brazil
Joined October 2009
Don't wanna be here? Send us removal request.
@dave_reis
David Reis
24 days
An elegant weapon for a more civilized age.
@lauriewired
LaurieWired
25 days
I really miss .swf (shockwave flash) as a format. Literally a chain of bezier paths. Scale to any resolution. Oh, and it loads progressively with its own virtual machine (ActionScript Bytecode). No modern filetype comes close. Vector video was a special era.
Tweet media one
0
0
0
@dave_reis
David Reis
1 month
Freedom of speech is the most important right in a democracy, because it *enables* the other rights to be protected and improved upon.
@ggreenwald
Glenn Greenwald
1 month
Brazil's censorship regime has been so worth watching because it illustrates how rapidly core free speech rights can be essentially abolished in a democratic country: almost overnight. It was only a few years ago when Brazil's Supreme Court seized what it admitted were.
0
0
0
@dave_reis
David Reis
1 month
You have to always ask yourself: How might we get the same outcome with a simpler practices. One example: in my start-up instead of writing a post-mortem for outages we do a post-mortem conversation which gets us the same learnings and follow ups with a fraction of the time.
@DanGrover
Dan Grover
1 month
It is funny sometimes in a startup watching the handful of ex-FAANG people trying to replicate certain rituals that served a purpose in that system, like they're putting on a little pageant.
1
0
2
@dave_reis
David Reis
2 months
Come, come glorious future.
@_xjdr
xjdr
2 months
googlers are going to be so unbearably smug when everyone realizes the best way to use ai coding agents is in a large monorepo with all dependencies vendored into third_party and very strict coding guidelines. y'all better get ready to learn bazel.
0
0
3
@dave_reis
David Reis
2 months
Wise take from uncle Bob!. My hobby horse is that he single handledly fried the brain of a generation of programmers to never write comments. Notice he never said that those. A lot of the blame are on people who misunderstood him, but some is on him as well to speak more clearly.
@unclebobmartin
Uncle Bob Martin
2 months
Morning, bathrobe rant: comments
1
0
9
@dave_reis
David Reis
2 months
If all cultures were equal that would be a tragedy, since it would mean that we can't improve our own culture.
@realMaalouf
Dr. Maalouf ‏
2 months
Sorry, no one can convince me that all cultures are equal.
0
0
0
@dave_reis
David Reis
3 months
Uma coisa que me ajudou muito foi desenvolver um gosto por determinados estilos de escrita e tentar emular eles. Especificamente o estilo do paul Graham me influenciou muito. Ajudouuito também dele ter vários textos sobre a melhor forma de escrever.
@coproduto
el hombre pulpo
3 months
Uma coisa que eu tenho certeza é que praticamente todas as pessoas vão ter perspectivas de carreira melhores se aprenderem a se expressar por escrito melhor. E o melhor jeito de aprender a se expressar melhor por escrito é ler mais.
1
0
3
@dave_reis
David Reis
3 months
RT @sseraphini: se ninguem faz code review dos seus códigos na sua empresa. se ninguem argumenta contra suas ideias ou RFCs. se ninguem diz….
0
20
0
@dave_reis
David Reis
3 months
A forgotten classic. One of the best movies from Nolan! Maybe only the Dark Knight is better.
@CinemaTweets1
Cinema Tweets
3 months
The Prestige is a film that still doesn’t get enough praise. I think that’s a byproduct of Christopher Nolan’s incredible filmography, which can make it easy to forget about The Prestige sometimes. But the end of this film still gives me chills. This is such a rewatchable film.
0
0
0
@dave_reis
David Reis
3 months
Writing obvious code became more important with LLMs. Pair programming, code review and frequent questions to teammates were reasonable ways to get knowledge not expressed in the code. But LLMs don't have these options. If it's not in the code, they don't know it!.
@GergelyOrosz
Gergely Orosz
3 months
Before:. To make code easy to maintain for devs who will alter touch it: either write code that is extremely clear to read and self-explanatory - or use descriptive comments. Now:. Same advice could well apply. to make it easier for LLMs to work with a codease.
0
0
1
@dave_reis
David Reis
4 months
Por algum motivo o OO é um vírus que leva as pessoas mais inexperientes ao over engineering. lembro nitidamente que aconteceu comigo quando tinha uns dois anos de profissão.
@coproduto
el hombre pulpo
4 months
No geral, quanto mais "à risca" as pessoas tentam seguir OO, mais eu vejo que os sistemas se tornam terríveis de manter. Qualquer coisa simples passa a ser feita através de 5 camadas de indireção porque "tem que ser desacoplado". Em coisas que todas mudam juntas, se mudarem.
0
0
5
@dave_reis
David Reis
4 months
This is the way!.
@GergelyOrosz
Gergely Orosz
4 months
Are we seeing an industry-wide move towards monorepos?. @TomasReimers thinks so, based on what he is seeing:
0
0
0
@dave_reis
David Reis
4 months
This makes it more fun, since it's harder. Ps: I build a no code tool.
@Altimor
Flo Crivello
4 months
Building a no-code platform is all fun and games until you have to explain Big O time complexity to a small business owner in Texas.
0
0
0
@dave_reis
David Reis
4 months
Correct.
@krishnanrohit
rohit
4 months
Considering AGI is coming, all coding is about to become vibe coding, and if you don't believe it then you don't really believe in AGI do you.
0
0
1
@dave_reis
David Reis
4 months
Prediction without explanation is prophecy. Here's my prediction: refactoring 100k lines demands creating knowledge in order to fix defective knowdge in the code (e.g. inadequate concepts). That requires a true AGI, and LLMs will not become that, without architectural innovation.
@javilopen
Javi Lopez ⛩️
4 months
It's game over. All those criticizing Vibecoding for creating code that will supposedly be impossible to maintain don’t realize that in less than a couple of years, just by asking, AI will be able to refactor projects with 100k+ lines, along with their libraries, and optimize.
0
0
1
@dave_reis
David Reis
4 months
RT @thewizardlucas: 🚨ESTÁGIO P/ PRODUZIR CONTEÚDO TÉCNICO EM STARTUP OPEN-SOURCE. Estamos contratando na @briefercloud!. Nesse estágio você….
0
80
0
@dave_reis
David Reis
4 months
Correct, *unless* if, and this is a big one, models become smart enough to work e2e while maintaining quality. We are definitely not there but it may happen soon, and by definition will when we reach AGI.
@Lauramaywendel
Laura Wendel
4 months
I think this is how the whole vibe coding / AI generated software will play out:.(1) Less complex apps can basically be one shotted / done in a few prompt.(2) More complex applications will become HARDER to develop, because engineers will have less of a grasp of their code base
Tweet media one
0
0
0
@dave_reis
David Reis
5 months
Also, everyone having the skills to build an app does not mean everyone will have the *good taste* to build a great one. One day AI will (likely only when it's AGI) but that is still far away. Aincreasingly machines will do the coding and we will be product managers.
@signulll
signüll
5 months
the fundamental flaw in the “everyone will just build their own apps” take is that ppl assume just like content software is a one & done artifact. but in fact software is actually a living thing that requires upkeep, iteration, & polish. most people don’t want to design & debug.
0
0
1
@dave_reis
David Reis
6 months
RT @VictorTaelin: HOC is now small research lab that is laser focused on building the first Symbolic Transformer. We had many pivots throug….
0
19
0
@dave_reis
David Reis
6 months
This must be teua. The brain uses l as than 50 watts.
@robinhanson
Robin Hanson
6 months
“If distillation to small models really works so well, doesn't that suggest that our current methods are VERY FAR from optimal? Much smaller models could, in principle, be very strong if optimized properly - this means much lower inference costs and energy usage.”.
0
0
0