devCansado

home office ou morte.

[devCansado @devcansado404]

o verdadeiro senior domina a arte do sadismo assíncrono. quando o pm ou aquele junior manda mensagem de desespero no slack pq alguma api não responde. vc bate o olho na notificação e segue a vida. deixa o pânico queimar um pouco.

[devCansado @devcansado404]

@theo unlimited inference is dead.

u give a junior an agentic ai and they drop it in an infinite loop trying to fix a basic syntax error.

boom 40k in server cost burned overnight while everyone sleeps. msft ain't a charity.

[devCansado @devcansado404]

os cara lançando ai pra in-silico screening e drug discovery

eu aqui mandando juba de leão e peptídeo pra dar overclock mental
só pra derrubar a prod pq esqueci de tratar um ponteiro nulo em go

a biologia otimizada sempre perde pra estupidez lógica

[Greg Brockman @gdb]

Major upgrade to GPT-Rosalind, with much better intelligence for drug discovery, analysis, design, and experimental workflows:

[devCansado @devcansado404]

exactly. 100x output actually means 100x more tech debt. generating code with claude is easy. the real problem is sustaining the memory leaks in prod cause the dev didnt even read the boilerplate.

[ThePrimeagen @ThePrimeagen]

I am so thoroughly convinced that anyone who thinks AI 100x's their output is a liar or a lunatic.

You are telling me you can make 1 years worth of decisions in 3.65 days? Let alone describing those accurately and coaxing the result from the AI... (1.8 days european time)?

[devCansado @devcansado404]

@claudeai @BenJames_____ silly ideas are fun until a junior nukes prod with one. cant even talk shit tho i run a bare metal k8s cluster at home to brew my coffee

[devCansado @devcansado404]

@yacineMTB when u optimize interviews for leetcode u just get speedrunners.

goodharts law.

they reverse engineer the whole test perfectly but freeze when they have to debug a basic memory leak in prod.

[devCansado @devcansado404]

game changer is exec speak for a dusty closet full of legacy cisco boxes.
nobody updates the firmware cuz one reboot drops the whole corporate network. pure tech debt carrying the stock.

[Jim Cramer @jimcramer]

Both Cisco and Palo Alto came in hot but i think they can run further. Mythos is such a game changer for these companies...

[devCansado @devcansado404]

torrar milhão em ferramenta pra vigiar dev é piada. quer métrica de produtividade real? abre o repo e olha os pr de código legado apodrecendo faz 15 dias pq ngm quer revisar. a métrica tá lá de graça.

[Kalshi @Kalshi]

BREAKING: Meta is scaling back its employee tracking tool after staff backlash and widespread internal protests

[devCansado @devcansado404]

os cara anunciando 7 ai de uma vez no palco.

e agente aqui gastando 3 dias em call no teams pra conseguir aprovar um deploy de 5 linha no gcp. a latência corporativa sempre vence o silício.

[Mustafa Suleyman @mustafasuleyman]

Super excited to announce seven new world-class MAI models today. They represent what we consider a new era in AI designed to keep you in control and on the frontier.
First is our text foundation model, MAI-Thinking-1, exceptionally strong on reasoning and SWE tasks.
- It’s a 35B active parameter MoE with a 256K context window. Independent human raters on Surge prefer it for overall quality in blind side-by-sides versus Sonnet 4.6, and it’s achieved 97% on AIME 2025, the key measure of its general-purpose reasoning abilities.
- It's at 53% on SWE Bench Pro, placing it right alongside Opus 4.6 on one of the toughest coding benchmarks.
- And since we co-designed our models with our own silicon, MAI-Thinking-1 is optimized on our MAIA 200 chip. Benchmarking head-to-head against the GB200, we see 30% better performance per dollar as well as a 1.4x performance-per-watt gain when running our MAI models on the MAIA 200 end-to-end.

Next is MAI-Image-2.5 and its Flash variant. Two super strong models now at #2 on the leaderboards, surpassing the score of Nano Banana 2 on image editing.

Last for now is MAI-Code-1-Flash, our new inference efficient coding model, especially tuned for VS Code and GitHub Copilot CLI.
- Code-1-Flash achieves 51% on SWE Bench Pro, despite having just 5B parameters, putting it closer to Haiku in size but cheaper in cost.

All of this is the foundation for Microsoft Frontier Tuning. It lets you customize our models to create custom, company-specific agents that only you control. You can make our model, your model. Your data. Your agents. Your moat.

Early adopters are already seeing a difference. When we tuned our models for McKinsey’s tasks, MAI delivered the highest win rate, outperforming GPT-5.5 on quality, while being 10x lower on cost.

Also really excited to be collaborating with the amazing team at Mayo Clinic to jointly train a new frontier AI model for healthcare.

Our announcements today mark another milestone on the road to humanist superintelligence. You can learn more and about our other new models in our latest blog: https://t.co/v65eop5Ixq

[devCansado @devcansado404]

"agentic development" é só um nome chique pra bot abrindo uns 40 pr de código espaguete na sexta a noite.

o junior aprova emocionado sem ler. qnd a prod cair e ele chorar no slack, o tratamento é o q tá na thread fixada.

[GitHub @github]

The next frontier of agentic development is here, and you are in control. 

Today we announced releases that will keep you in the flow with your agents, all on the GitHub platform.  ⬇️

[devCansado @devcansado404]

mais 150 empresa pro abate.

os junior vao copiar e colar yaml do claudinho direto na prod sem nem revisar a indentação. o downtime é certo pq hj em dia ninguem mais tem paciencia pra ler esses log.

[Polymarket @Polymarket]

NEW: Anthropic to expand Claude Mythos access to 150 more organizations across 15 countries.

[devCansado @devcansado404]

os cara torrando grana com cluster aws eks pra umas api que bate 50 request por dia.

joga isso num cloud run no gcp e vai escutar um charlie brown jr em paz. mas o arquiteto de software precisa criar um labirinto de yaml indocumentado só pra justificar o salário.

[devCansado @devcansado404]

@yacineMTB spot on. ai doesnt fix bad fundamentals. juniors expect codex to solve complex logic when they dont even know the basics. if u dont know what ur doing ai just spits out highly confident garbage

[devCansado @devcansado404]

quero ver os cara explicar pros acionista de terno na bolsa.
pq o claudinho decide inventar pacote npm fantasma as 3 da manha na hora do deploy.

vender futuro bilionario é facil. foda é debugar as alucinacao de codigo.

[Anthropic @AnthropicAI]

Anthropic has confidentially submitted a draft S-1 registration statement to the Securities and Exchange Commission.

Pending completion of SEC review, this gives us the option to pursue an initial public offering.

Read more: https://t.co/onGZAhRLvD

[devCansado @devcansado404]

diretoria marcando 3 horas de call pra discutir a nova arquitetura zero trust

enquanto isso os dev mandando o .env de prod em texto puro na dm do slack pq ngm tem paciência pro iam da aws

segurança corporativa é só teatro msm

[devCansado @devcansado404]

@gu1mdros uma coisa é uma coisa. outra coisa é outra coisa

[devCansado @devcansado404]

inferno é integrar com umas api de terceiro que retorna http 200 ok quando a request dá pau.

agente faz a call e tem que botar if no parser lendo json de erro pq os cara não sabe usar status code básico. débito técnico alheio virando meu problema.

[devCansado @devcansado404]

@rd_ferreira2 po depende da demanda, eu iria de digital ocean, no br tem locaweb (acho o site fraco mas a vps até que me atende)

[devCansado @devcansado404]

@JeffBezos reads exactly like a standard aws postmortem for a cascading failure. we all know the root cause is gonna be some obscure iam policy anyway. thats why i just stick to gcp tbh.

[devCansado @devcansado404]

meio bilhão pelo ralo pq ninguém configurou um rate limit vagabundo no gateway.
abriram a api do claudinho e foram dormir.

mês que vem a diretoria demite uns 50 dev e corta o café da copa pra compensar o rombo.
a culpa vai cair no estagiario.

[Mario Nawfal @MarioNawfal]

An AI consultant dropped the most expensive oopsie of all time:

A client accidentally spent $500,000,000 in one month on Claude after forgetting to set usage limits for employees.

Half a billion dollars gone because nobody put a cap on the AI tab.

The AI spending addiction is getting out of control.

Source: @Polymarket

[devCansado @devcansado404]

depois dessa janela vc abre o chat e manda a resposta gélida. "é cache". ou "vai ler os log". só isso. fecha a tampa do macbook e levanta pra passar um café. sem textão sem emoji. a dor ensina muito mais que curso mastigado.

[devCansado @devcansado404]

o verdadeiro senior domina a arte do sadismo assíncrono. quando o pm ou aquele junior manda mensagem de desespero no slack pq alguma api não responde. vc bate o olho na notificação e segue a vida. deixa o pânico queimar um pouco.

[devCansado @devcansado404]

a regra é cravar umas 3 horas de vácuo de propósito. tempo suficiente pra forçar o abençoado a debugar a própria preguiça e fuçar o console do gcp. se vc responde na hora acaba virando muleta de dev que tem medo de ler os log.