Martín retweetledi
Que increíble, un ARGENTINO tiene que encargarse una vez más de ser el FARO de OCCIDENTE.
Owen Kemeys@OwenKemeys
We live in hell
Español
Martín
714 posts
Martín
@emebermz
paleo código
Ciudad Autónoma de Buenos Aire Katılım Haziran 2011
802 Takip Edilen65 Takipçiler
Micron $MU el 18/3 presentó resultados de su 2QFY26 superando ampliamente los consensos mas un guidance que la sacó de la cancha.
La acción tocó 471 dólares ese día marcando ATH y corrigió muy fuerte un 34% en 9 ruedas.
Desde ahí lleva un rebote del 46% en 9 ruedas y está a un 3.5% de su ATH.
A estos precios $MU cotiza a un Forward P/E de 7x después de haber subido "apenas" un 59% YTD y 540% en 12 meses.
De locos.
Español
Martín retweetledi
PewDiePie trained a frontier model at home and beat OpenAI and Gemini
> be me PewDiePie
> play games on YouTube and scream 24/7
> become a meme reviewer
> get unfathomably famous
> "fuck that" I'm a family guy now
> retire
> move to Japan with my beautiful wife
> mfw I'm a dad now
> "as a dad I must do dad things"
> scratch that
> "as a dad I must do frontier AI research"
> goal is to beat GPT-4o at coding (16% on Aider)
> buys $30000 GPU setup
> reads DeepSeek paper
> decides to start massive GitHub scraping and data augmenting run
> not good enough
> read Magicoder paper
> generate tons of synthetic coding data
> train a new model
> guuuuuh. the data made the model worse
> mfw I just wasted months for nothing
> decides to lock in and try again
> makes model worse again ffs
> try again
> finally beating GPT-4o on data (16.1%)
> not satisfied
> "I should simply train a reasoning model"
> reads more papers
> start experimenting with more synthetic data
> "Mhhh something doesn't smell right"
> house almost burned down due to power connector
> shrug
> just buy a new one
> mfw computer is now crashing 24/7 generating synthetic data
> new plan: just call DeepSeek API for high quality synthetic data
> train model again
> 17.2%
> performance fluctuates slightly on each eval run
> big brain idea: repeat eval until we randomly reach >18%
> sike actually got 19.6%
> feelsgoodman.png
> nvm the benchmark was contaminated and I was training the wrong base model the whole time
> rerun everything again
> new score: 4.4%
> you read that right REEEEEEEE
> almost get a heart attack
> "have you tried plugging the device off and back on again?"
> change nothing and just retrain again
> 25.3 %
> LETS FUCKING GOOO
> realize that 1/3rd of the benchmark was not running. guuuh
> scared shitless it would score below 10% again
> run yet again. the whole thing this time
> Thirty fucking six percent
> accidentally beat Gemini 2.0 Pro Exp and GPT-4.1 mini
> pops the AI bubble
> "I want moaaaar"
> finds some more post-training data
> 39% babyyyy
> realize at the end that I was just benchmaxxing Aider polyglot
> next quest: run SWE-Bench and other coding benchmarks
> "I failed a thousand times, but prevailed in the end"
> just a little sad side-quest
> probably going to train GPT-6 myself by next month
English
Tuvo que poner candado porque lo que se venía era excesivo
Español
Martín retweetledi
It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but specifically this last December. There are a number of asterisks but imo coding agents basically didn’t work before December and basically work since - the models have significantly higher quality, long-term coherence and tenacity and they can power through large and long tasks, well past enough that it is extremely disruptive to the default programming workflow.
Just to give an example, over the weekend I was building a local video analysis dashboard for the cameras of my home so I wrote: “Here is the local IP and username/password of my DGX Spark. Log in, set up ssh keys, set up vLLM, download and bench Qwen3-VL, set up a server endpoint to inference videos, a basic web ui dashboard, test everything, set it up with systemd, record memory notes for yourself and write up a markdown report for me”. The agent went off for ~30 minutes, ran into multiple issues, researched solutions online, resolved them one by one, wrote the code, tested it, debugged it, set up the services, and came back with the report and it was just done. I didn’t touch anything. All of this could easily have been a weekend project just 3 months ago but today it’s something you kick off and forget about for 30 minutes.
As a result, programming is becoming unrecognizable. You’re not typing computer code into an editor like the way things were since computers were invented, that era is over. You're spinning up AI agents, giving them tasks *in English* and managing and reviewing their work in parallel. The biggest prize is in figuring out how you can keep ascending the layers of abstraction to set up long-running orchestrator Claws with all of the right tools, memory and instructions that productively manage multiple parallel Code instances for you. The leverage achievable via top tier "agentic engineering" feels very high right now.
It’s not perfect, it needs high-level direction, judgement, taste, oversight, iteration and hints and ideas. It works a lot better in some scenarios than others (e.g. especially for tasks that are well-specified and where you can verify/test functionality). The key is to build intuition to decompose the task just right to hand off the parts that work and help out around the edges. But imo, this is nowhere near "business as usual" time in software.
English
Martín retweetledi
We built a browser with GPT-5.2 in Cursor. It ran uninterrupted for one week.
It's 3M+ lines of code across thousands of files. The rendering engine is from-scratch in Rust with HTML parsing, CSS cascade, layout, text shaping, paint, and a custom JS VM.
It *kind of* works! It still has issues and is of course very far from Webkit/Chromium parity, but we were astonished that simple websites render quickly and largely correctly.
Cursor@cursor_ai
GPT-5.2 Codex is now available in Cursor! We believe it's the frontier model for long-running tasks.
English
Martín retweetledi
Martín retweetledi
I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse and between. I have a sense that I could be 10X more powerful if I just properly string together what has become available over the last ~year and a failure to claim the boost feels decidedly like skill issue. There's a new programmable layer of abstraction to master (in addition to the usual layers below) involving agents, subagents, their prompts, contexts, memory, modes, permissions, tools, plugins, skills, hooks, MCP, LSP, slash commands, workflows, IDE integrations, and a need to build an all-encompassing mental model for strengths and pitfalls of fundamentally stochastic, fallible, unintelligible and changing entities suddenly intermingled with what used to be good old fashioned engineering. Clearly some powerful alien tool was handed around except it comes with no manual and everyone has to figure out how to hold it and operate it, while the resulting magnitude 9 earthquake is rocking the profession. Roll up your sleeves to not fall behind.
English
Martín retweetledi
CAMBIOS DE HORARIOS POR CONDICIONES CLIMÁTICAS ADVERSAS
Debido a las condiciones climáticas previstas por el Servicio Meteorológico para la medianoche del sábado 15 de noviembre, el show de Oasis, originalmente programado para las 21:00 hs, adelantará su inicio a las 20:30 hs.
Por este motivo, los horarios para el show del 15/11 serán los siguientes:
Puertas: 16:00 hs
Richard Ashcroft: 19:15 hs
Oasis: 20:30 hs
Esta medida busca priorizar la seguridad y el bienestar de todos los asistentes, asegurando que puedan regresar a sus hogares de manera segura.
Les recomendamos mantenerse atentos ante cualquier actualización sobre el clima en las redes sociales de DF Entertainment y en allaccess.com.ar
Español
