I’ve got 2 free months of LinkedIn Premium to give away from LinkedIn’s referral program.
If you’re job hunting, building your business, or figuring out your next move drop a comment or DM me. Happy to share!
#LinkedInPremium#PayItForward
I’ve got 2 free months of LinkedIn Premium to give away from LinkedIn’s referral program.
If you’re job hunting, building your business, or figuring out your next move drop a comment or DM me. Happy to share!
#LinkedInPremium#PayItForward
3 tips to save context space:
1. Vague prompts burn more tokens.
2. Manage MCP servers. They load every tool by default. Turn off what you don’t need. Use Skills instead.
3. Use subagents. They work in their own context & only return summary. Main agent stays clean.
3 critical lessons from (Agentic Loop):
1. stop_reason matters
end_turn = done | tool_use = run tool | max_tokens = cut off
2. Model has zero memory.
History is everything.
Miss a result → broken agent or infinite loop.
3. Always set max iterations.
Or costs explode.
Day 1: Agentic AI + orchestration.
Stopped debating “Will AI replace devs?”
Started learning..
Agentic AI ≠ Prompt → Response
It’s: Think → Act → Observe → Decide → Repeat
Biggest lesson:
Building an agent isn’t calling an LLM.
It’s building the loop around it.