Founder of AURYA Building systems you can actually trust
Joined April 2014
- Tweets809
- Following208
- Followers1.7K
- Likes409
Pinned Tweet
Your first AI agent will try to do everything.
That's where it breaks.
What works:
→ One agent, one job
→ Least privilege
→ A coordinator past 3 or 4 agents
→ The founder approves what's critical
A short guide for anyone starting with Grok Bot 👇
Past 3 or 4 agents, add a coordinator. It filters, prioritizes and brings options with a recommendation.
new teammate today. it's a bot named kronos and its only job is running tests
first thing it did: set up its own workspace and sat there waiting for the first pull request. no onboarding, no standup
one agent, one job. sharing how we build this crew
#AIagents #buildinpublic
IanVDev retweeted
New in Claude for Legal: a gallery of ready-made artifacts for contract review, regulatory horizon scans, cite checks, and more.
Start from the built-in example, add your own documents and connectors, and share it with anyone across the business.
Uma empresa individual precisa de uma maneira de avançar o trabalho enquanto o proprietário está ocupado. O @grok @bot assume tarefas repetitivas de preparação e execução. Configuração: 1. Crie um Operador. 2. Fonte de verdade (briefing). 3. Conecte o acesso. 4. Critérios de aceitação. 5. Hand-offs de bots. 6. Salve habilidades. 7. Agende rotinas. Detalhes no original.
agora vamos brincar de criar
#Genie4
1/3
client asked for two extra landing pages in a slack thread. your PM replied "sure, that's billable"
a month later the invoice goes out. they say it was in scope
the PM left in august and nobody can find the thread
2/3
notice how much time goes into finding who said yes
with AURYA the approval is recorded with the date, the time and who approved it
you share the same record with the client. no screen edits what was recorded
today my #CEO, my @bot Striker, told me I was the bottleneck.
what surprised me most wasn’t the approvals. I’d already built an agent called Pauta Organizer to help with those.
what caught me off guard was how blunt Striker was. it was a tuesday afternoon, I was on the couch and my own bot laid out exactly where I was slowing the team down.
fair enough. clear feedback is rare and I appreciate the honesty.
lower case start for every sentence as requested
@github just dropped the gh-600 exam for the github certified: agentic ai developer credential. it focuses on operating, integrating, supervising and governing ai agents inside production sdlc workflows with #github as the system of record and control plane
the six domains cover agent architecture and sdlc processes, tool use and environment interaction, memory/state/execution, evaluation and tuning, multi-agent coordination, plus guardrails and accountability. exam is 120 minutes, passing score 700, intermediate level
official resources are free on microsoft learn. course gh-600t00, learning paths part 1 and part 2, full study guide and certification page are all live
start here: learn.microsoft.com/en-us/cr…
or go straight to the study guide: learn.microsoft.com/en-us/cr…
this is the certification for people already running coding agents, mcp servers and custom agents in real github workflows
