peteralbert

3 ms·

1.
How to cut the cost of long AI agent threads (without making the agent dumber)(viktor.com)
2 points by peteralbert 4mo ago|0 comments
2.
by peteralbert 7mo ago
Our agent is prompted to update the skill files, whenever it encounters errors, or sees something outdated. We also have a seperate clean up agent on a cron that every week cleans up skill files, deduplicates info, and deletes unecessary de
3.
What we learned building an AI agent with 100K tools in production(getviktor.com)
2 points by peteralbert 7mo ago|0 comments
4.
by peteralbert 7mo ago
Author here (Peter, CTO). We've been iterating on this for about a year. Happy to answer questions on any section.
5.
Giving our AI agent 100k tools made it worse(getviktor.com)
1 points by peteralbert 7mo ago|1 comments
6.
by peteralbert 7mo ago
Author here (Peter, CTO). We've been iterating on this for about a year. Happy to answer questions on any section.
7.
Why LLM agents break when you give them 100k tools(getviktor.com)
4 points by peteralbert 7mo ago|3 comments
8.
by peteralbert 7mo ago
We validate tool calls with Pydantic models built directly from the JSON schemas, running inside the container. So the agent gets instant feedback if it passes the wrong type, misses a required field, or hallucinates a parameter — before an
9.
by peteralbert 7mo ago
Author here (Peter, CTO). We've been iterating on this for about a year. Happy to answer questions on any section.
10.
What breaks when your agent has 100k tools(getviktor.com)
11 points by peteralbert 7mo ago|5 comments