Skip to main content

llm

14 articles

How do you build AI agents that learn from expert feedback?
Guide evaluation ai-agents

How do you build AI agents that learn from expert feedback?

Build agents that learn from experts by logging pass/fail verdicts with reasons, turning them into tests, changing prompts or memory, and re-running the tests.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 8 min read
How do you build AI agents inside your own cloud and model approval process?
Guide ai-agents enterprise-ai

How do you build AI agents inside your own cloud and model approval process?

Build agents on the model service your cloud team approved, keep traffic private, check retention per model, log every call, and test before widening access.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 7 min read
How do you fix inconsistent LLM output in a review workflow?
Guide evaluation llm

How do you fix inconsistent LLM output in a review workflow?

Fix inconsistent LLM output by locking the output format, writing a binary expert rubric and measuring agreement. Temperature 0 alone does not fix it.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 6 min read
How many labeled examples do you need before you trust an LLM judge?
Guide evaluation llm

How many labeled examples do you need before you trust an LLM judge?

No single number exists: practitioners suggest about 100 expert labels per failure mode, balanced Pass and Fail, then a confidence interval to check trust.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 7 min read
Which LLM eval tools do practitioners use, and where do they fall short?
Guide evaluation llm

Which LLM eval tools do practitioners use, and where do they fall short?

Evals expert Hamel Husain meets LangSmith, Arize and Braintrust most often, but no eval tool supplies your rubric, expert labels or error analysis for you.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 6 min read
How do you measure whether your AI agent is getting better each sprint?
Guide evaluation ai-agents

How do you measure whether your AI agent is getting better each sprint?

Run the same fixed test set every sprint, track pass rates per failure type with confidence intervals, and use a regression suite so fixes keep old wins.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 6 min read
When should you compare two AI agent versions side by side instead of scoring one output?
Guide evaluation llm

When should you compare two AI agent versions side by side instead of scoring one output?

Compare two agent versions side by side for subjective qualities like tone; score one output for objective checks like faithfulness. Always swap answer order.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 7 min read
How do you turn expert feedback on AI outputs into a grading rubric?
Guide evaluation llm

How do you turn expert feedback on AI outputs into a grading rubric?

Have one expert write notes on real outputs, group the notes into failure types, turn each into a pass/fail check, then test the rubric against fresh labels.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 7 min read
LLM-as-Judge: Using AI to Grade AI
Guide evaluation LLM

LLM-as-Judge: Using AI to Grade AI

Hamel Husain's critique shadowing approach checks an LLM judge against a domain expert's labels. Here is the practical guide for AI product teams.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 12 min read
Context Windows Are Not Memory
Opinion ai-agents memory

Context Windows Are Not Memory

Your AI agent has a 128K context window. You think it remembers. It does not. Context windows are temporary buffers, not persistent understanding. Here is why the distinction matters and what to do about it.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 10 min read
Why Your AI Chatbot Sounds Generic After 100 Conversations
Deep Dive LLM personalization

Why Your AI Chatbot Sounds Generic After 100 Conversations

Your AI chatbot had 100 conversations and still treats every user like a stranger. The problem is not the model, it is that your AI has amnesia by design.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 13 min read
Why Your LLM Wrapper Is a Commodity (And What to Do About It)
Opinion LLM architecture

Why Your LLM Wrapper Is a Commodity (And What to Do About It)

If your product is a thin wrapper around an LLM API, you are one model update away from irrelevance. The moat is not in the model. It is in the user-level intelligence layer you build on top.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 11 min read
RAG Is Not Personalization
Deep Dive RAG personalization

RAG Is Not Personalization

Retrieval-augmented generation gives AI products better answers, not personal ones. True personalization requires self-models that understand the user, not just the query.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 16 min read
Why Your LLM Sounds Generic (And How Self-Models Fix It)
Opinion LLM self-models

Why Your LLM Sounds Generic (And How Self-Models Fix It)

Your LLM doesn't have a quality problem. It has an identity problem. It doesn't know who it's talking to. Self-models give every AI interaction a 'who am I talking to' layer.
Robert Ta's Self-Model
Robert Ta's Self-Model
· 7 min read

Stay sharp on AI personalization

Daily insights and research on AI personalization and context management at scale. Read by hundreds of AI builders.

Daily articles on AI-native products. Unsubscribe anytime.

Building AI that needs to understand its users?

Discuss a scoped API pilot