Gosto de Café, PHP e crio galinhas! Building open Source DotCrow.si Outros links no link abaixo 👇🏻
Buriticupu, Brazil 🇧🇷
- Tweets 3,240
- Following 702
- Followers 3,928
- Likes 2,915
Pinned Tweet
Li o paper. De fato o 65x é real, mas é contra um método de 2019 que nem RAG é. O que muda na prática?
RAG is officially cooked.
IBM built a retriever that hallucinates 65x less than fine-tuned RAG systems.
Right now, every company is using Retrieval-Augmented Generation (RAG) to stop their AI from hallucinating.
The idea is to give the AI a private database to read so it doesn't make things up.
But there is a massive structural flaw.
Standard RAG blindly grabs documents based on vector similarity and forces the AI to read them. If the retriever pulls irrelevant garbage, the AI doesn't stop.
It just hallucinates with citations.
IBM just dropped a paper that changes the math entirely.
They built a new retrieval architecture that hallucinates 65x less than state-of-the-art fine-tuned RAG systems.
65x is the difference between a fun prototype and a production-grade enterprise system.
Instead of blindly dumping context into the model and hoping for the best, their system actively filters and validates the retrieved data before the generation step even begins.
It cuts out the noise. It forces absolute precision.
For businesses, this is the holy grail.
Zero hallucinations means zero liability. It means autonomous agents can finally be trusted with sensitive company data without a human babysitting every output.
4
3
443
Li o paper. De fato o 65x é real, mas é contra um método de 2019 que nem RAG é. O que muda na prática?
RAG is officially cooked.
IBM built a retriever that hallucinates 65x less than fine-tuned RAG systems.
Right now, every company is using Retrieval-Augmented Generation (RAG) to stop their AI from hallucinating.
The idea is to give the AI a private database to read so it doesn't make things up.
But there is a massive structural flaw.
Standard RAG blindly grabs documents based on vector similarity and forces the AI to read them. If the retriever pulls irrelevant garbage, the AI doesn't stop.
It just hallucinates with citations.
IBM just dropped a paper that changes the math entirely.
They built a new retrieval architecture that hallucinates 65x less than state-of-the-art fine-tuned RAG systems.
65x is the difference between a fun prototype and a production-grade enterprise system.
Instead of blindly dumping context into the model and hoping for the best, their system actively filters and validates the retrieved data before the generation step even begins.
It cuts out the noise. It forces absolute precision.
For businesses, this is the holy grail.
Zero hallucinations means zero liability. It means autonomous agents can finally be trusted with sensitive company data without a human babysitting every output.
4
3
443
Inserir um lembrete para meu Espaço! x.com/i/spaces/1YGNrblRrgnGw
133
Link do BetaLab no meu perfil. Lá posto as novidades antes de todo mundo ou um pouco depois de todo mundo haha Crio galinhas também! Kkkkk Então me dá um descanso kkkkkkk
58
Hoje às 19h a gente abre esse paper ao vivo no espaço, vou mandar o PDF no Beta Lab. No espaço de hoje vou mostrar onde a arquitetura realmente inova. Te espero lá!
1
22
Passei uns dias fora mas estou oficialmente de volta, façam perguntas da bolha tech, vou responder todas em um vídeo.
1
282
Posso considerar nossos dados Open Source?
🚨 148 Million Brazilian PII Records Exposed
A misconfigured Elasticsearch instance was discovered, potentially exposing over 148 million records containing Brazilian citizens' personally identifiable information (PII).
The exposed data includes:
- CPF (Brazilian taxpayer ID) numbers
- Full names
- Dates of birth
- Residential addresses and postal codes
- Phone numbers
Validation: Sample records were cross-checked against Brazil’s official CPF registry, confirming matching identity details.
The dataset shares similarities with previously reported Brazilian data leaks, including the massive 2021 breach. However, its original source remains unconfirmed. Reported to the hosting provider.
⚠️ Risks: Identity theft, financial fraud, targeted phishing, and social engineering.
#ThreatIntel #CyberSecurity #DataBreach
4
378
Com base nessa enquete, vcs tem plano de sair do Brasil ou vão buscar formas de continuar aqui? Se sim pq? Quero ouvir algumas opiniões pra tomar algumas decisões.
2
1
311
Novembro:
dobrar a aposta no que já tá funcionando
ou
testar algo totalmente novo antes do ano virar?
1
1
198
Opinião depois de 1 mês de build in public: mostrar o processo não atrai só seguidor.
Atrai cliente, sócio, concorrente e crítico.
Vale a pena mesmo assim. Esconder a ideia nunca protegeu ninguém.
11
1
48
1,403
Ontem foi o primeiro evento do Beta Lab First e só tenho a agradecer pela troca de conhecimentos.
1
119
Coloquei scripts no DotCrow que anonimizam dados sensíveis antes de qualquer coisa ir pro modelo de IA escolhido.
Open source é isso: você vê exatamente o que sai do seu computador.
Agora uma pergunta, você confia em mandar seus arquivos pra uma IA fechada?
2
5
355
Boa tarde Founders!
Os e-mails e invites já foram enviados a todos os palestrantes e Founders com Pitch!
☕️Todos os ouvintes também já receberam o e-mail com o link de nossa sala para hoje às 19h.
🎟️ Ouvintes com algum projeto ou ideia ganham R$ 30 reais de desconto no ingresso.
CORRE QUE AINDA DÁ TEMPO!
Link abaixo!
🚨🚨3 vagas restantes para Founders Pitch🚨🚨
💲💰Ganhe 30% do valor da inscrição indicando alguém para este evento!
LINK ABAIXO
👇🏻
1
1
399
Daqui a pouco vou tentar ver o código fonte
🚨 850+ compromised GitHub accounts. 13,500+ malicious branches.
Researchers uncovered 'Gheist,' an ongoing campaign using stolen tokens to plant GitHub Actions workflows that dump private repo secrets into logs.
Attackers download the logs, then delete the runs.
Details: thehackernews.com/2026/10/gh…
4
396
Você está feliz ou triste com a vitória de @LulaOficial
6%
Muito feliz
9%
Feliz
15%
Triste
71%
Muito triste
34 votes • Final results
2
540