Bartosz Mikulski
@mikulskibartosz.bsky.social
Stop AI Hallucinations in Fintech
Risk Audits + Custom Roadmaps for customer-facing AI
☑️ Creator: laoshu.ai OSS citation checker
📧 Hallucination fire-drill → mikulskibartosz.name
Risk Audits + Custom Roadmaps for customer-facing AI
☑️ Creator: laoshu.ai OSS citation checker
📧 Hallucination fire-drill → mikulskibartosz.name
Your RAG hallucinates. Here's why.
By the time you're tweaking prompts, it's already too late.
I've seen teams spend months optimizing prompts and retrieval parameters, trying to stop hallucinations. But the real issue started weeks before they wrote the first line of their RAG code.
By the time you're tweaking prompts, it's already too late.
I've seen teams spend months optimizing prompts and retrieval parameters, trying to stop hallucinations. But the real issue started weeks before they wrote the first line of their RAG code.
July 15, 2025 at 2:31 AM
Your RAG hallucinates. Here's why.
By the time you're tweaking prompts, it's already too late.
I've seen teams spend months optimizing prompts and retrieval parameters, trying to stop hallucinations. But the real issue started weeks before they wrote the first line of their RAG code.
By the time you're tweaking prompts, it's already too late.
I've seen teams spend months optimizing prompts and retrieval parameters, trying to stop hallucinations. But the real issue started weeks before they wrote the first line of their RAG code.
Your AI is lying
Most fintech leaders think hallucinations are a model problem. They’re not.
Hallucinations are a debugging problem.
Jason Liu and Hamel Husain both agree: most AI projects succeed (or fail) because of how they handle evaluation.
Most fintech leaders think hallucinations are a model problem. They’re not.
Hallucinations are a debugging problem.
Jason Liu and Hamel Husain both agree: most AI projects succeed (or fail) because of how they handle evaluation.
July 14, 2025 at 5:26 AM
Your AI is lying
Most fintech leaders think hallucinations are a model problem. They’re not.
Hallucinations are a debugging problem.
Jason Liu and Hamel Husain both agree: most AI projects succeed (or fail) because of how they handle evaluation.
Most fintech leaders think hallucinations are a model problem. They’re not.
Hallucinations are a debugging problem.
Jason Liu and Hamel Husain both agree: most AI projects succeed (or fail) because of how they handle evaluation.
LLMs bluff. Your reputation bleeds. That's why I'm building Laoshu 🐭.
It can automatically verify the citations in Perplexity output, and it's a free, open-source project.
It can automatically verify the citations in Perplexity output, and it's a free, open-source project.
July 11, 2025 at 1:25 PM
LLMs bluff. Your reputation bleeds. That's why I'm building Laoshu 🐭.
It can automatically verify the citations in Perplexity output, and it's a free, open-source project.
It can automatically verify the citations in Perplexity output, and it's a free, open-source project.