· 4 min read
Small Models, Big Inbox: Where 200M-Param MoE Actually Wins
A recent 200M-parameter hobbyist MoE is a great reminder: email triage, classification, and routing rarely need frontier models. Here's where small wins.
All#ai#ai-agents#ai-productivity#applied-ai#automation#calendar#context-switching#deep-work#documents#email#engineering#engineering-leadership#invisible-ai#knowledge-work#llm-evaluation#product-updates#productivity#prompt-engineering#remote-work#search#security#spreadsheets#startups#storage#tamaton#unified-workspace#workflow#workflows
· 4 min read
RAG Is Not Search, and Your Inbox Proves It
Retrieval-augmented generation quietly breaks on email and files because it ranks documents, not conversations. Here's why, and what to do about it.
· 4 min read
The GRASP Problem: Why General LLM Benchmarks Lie to You
Leaderboard scores predict almost nothing about how a model performs on your actual domain tasks. The GRASP paradigm shows why — and what to measure instead.
· 4 min read
Give Your Calendar an Agent, Not Another Assistant Bot
A scheduling bot books slots. A real AI calendar agent remembers, decides, and negotiates. Here's how to design the second one.
· 4 min read
AI Agents vs. Assistants: What Actually Changes for Your Work
The real difference between AI agents and assistants isn't intelligence — it's who owns the loop from goal to done, and what that demands from your tools.
· 4 min read
Choosing an LLM for Real Work, Not Benchmarks
A practical framework for picking the best AI model for productivity tasks — weighing latency, cost, and reliability over leaderboard scores.
· 4 min read
Evaluating LLM Output Without Fooling Yourself
Vibes-based testing is why AI features break in production. Here's how to build cheap, repeatable evals for summaries, extraction, and drafts.
· 4 min read
How to Evaluate a RAG System Before You Trust It
A concrete evaluation playbook for retrieval recall, grounding, citation faithfulness, and the failure modes that quietly wreck RAG systems.
· 4 min read
The Agent Reliability Gap: Why 90% Isn't Good Enough
Multi-step agents compound their error rate — a 95%-accurate step becomes a coin flip after 15 actions. Here's the math and how to design around it.
· 4 min read
How to Evaluate an AI Feature When 'It Feels Better'
A practical eval playbook for AI-assisted knowledge work: build golden sets, run LLM-as-judge checks, and catch regressions without a research team.
· 4 min read
One Agent or Many? A Sober Take on Multi-Agent Hype
Multi-agent systems are trending, but most productivity work is better served by one well-scoped agent with good tools. Here's when orchestration actually earns its keep.
· 4 min read
HalluSquatting: When Your Coding Agent Trusts a Fake Repo
Hallucinated package names are becoming a real supply-chain attack. Here's how HalluSquatting works against AI coding agents — and how to defend against it.
Get started
Reserve your @tamaton.ai email
Claim your address before someone else does — free to start, with an AI-native inbox built in.