
Rafael Lima de Souza
4K posts

Rafael Lima de Souza
@lima_rafael
Programmer / Game Designer at @BananaBytesBr, Coordenador Pedagógico na Your Course, I created Sophstar e eu tweeto in portuguese and em inglês.
















Uniracers is a technical marvel, maintaining its speed when most other SNES games would start chugging. It's also a fascinating story. In 1994 shortly after the game launched, Pixar of all companies sued DMA Design, claiming the Unicycles ripped off their 1987 short Red's Dream. Pixar ultimately won (or the publisher, Nintendo settled) and the game was ordered to cease production after only 300,000 copies were made. Pixar had a point, the Unicycles looked similar. They're also unicycles. Once you animate those, there's not many differences that will happen. Interestingly enough, this whole legal debacle was a shift in direction for the Lemmings developer. Uniracers was their pitch and Nintendo loved it. If Pixar hadn't pulled the rug out from under them, they would have their own pillar under Nintendo's flag. Without Uniracers, DMA had to start over without their family-friendly franchise. Nintendo wanted them on the Ultra 64, but it soon became clear to both companies that only one of them was interested in making a cute N64 mascot. Body Harvest is a little tale on its own, but the funniest part of the story is how DMA Design started working on this small PC game called Race n' Chase at the same time. DMA Design released Grand Theft Auto in 1997, rebranding into Rockstar North after releasing GTA 3. If Pixar hadn't done a Disney, we might be Uniracing in Mario Kart World right now. Meanwhile #GrandTheftAuto might be nothing more than a spark in David Jones' eye. Probably not, but it's fascinating speculation. Also this game is dope.













AMD Senior AI Director confirms Claude has been nerfed. She analyzed Claude's session logs from Janurary to March: > median thinking dropped from ~2,200 to ~600 chars > API requests went up 80x from Feb to Mar. less thinking and failed attempts meaning more retries, burning more tokens, and spending more on tokens > reads-per-edit dropped from 6.6x → 2.0x. model stops researching code before touching it. > model tried to bail out or ask "should i continue" 173 times in 17 days (0 times before March 8). > self-contradiction in reasoning ("oh wait, actually...") tripled. > conventions like CLAUDE.md get ignored because there's less thinking budget to cross-check edits > 5pm and 7pm PST are the worst hours, late night is significantly better. this means the thinking allocation is most likely GPU-load-sensitive.


Esse tipo de pergunta faz sentido pra um concurso público? Ele avalia o que de útil para o dia a dia ? (mas qual a resposta certa?)








