Rod Rivera | AI Automations
@rodriveraai.bsky.social
Helping you become an AI Operator & Automator
💡 AI hot takes, news, explainers
🗯️ Fun educational AI comics
Follow for daily AI tips and news!
💡 AI hot takes, news, explainers
🗯️ Fun educational AI comics
Follow for daily AI tips and news!
7/ If you’re building, researching, or just curious about agents…
📅 Mark your calendar: October 15th
📍 London
RSVP: www.meetup.com/london-agent...
See you there! 👋
📅 Mark your calendar: October 15th
📍 London
RSVP: www.meetup.com/london-agent...
See you there! 👋
Agentic AI-MeetUp London #2 Wednesday 15th October, Wed, Oct 15, 2025, 6:00 PM | Meetup
# London Agentic AI Meet-Up #2 - October **Hosted by V7Labs.** **Register now to secure your spot — spaces are limited and will go fast!** Join us on Wednesday, 15th Oct
www.meetup.com
October 1, 2025 at 8:54 AM
7/ If you’re building, researching, or just curious about agents…
📅 Mark your calendar: October 15th
📍 London
RSVP: www.meetup.com/london-agent...
See you there! 👋
📅 Mark your calendar: October 15th
📍 London
RSVP: www.meetup.com/london-agent...
See you there! 👋
6/ My takeaways:
1. Choose your agent patterns carefully
2. Prioritize evaluation frameworks
3. It's not about replacing developers. It's about amplifying what they can build.
1. Choose your agent patterns carefully
2. Prioritize evaluation frameworks
3. It's not about replacing developers. It's about amplifying what they can build.
Agentic AI-MeetUp London #2 Wednesday 15th October, Wed, Oct 15, 2025, 6:00 PM | Meetup
# London Agentic AI Meet-Up #2 - October **Hosted by V7Labs.** **Register now to secure your spot — spaces are limited and will go fast!** Join us on Wednesday, 15th Oct
www.meetup.com
October 1, 2025 at 8:54 AM
6/ My takeaways:
1. Choose your agent patterns carefully
2. Prioritize evaluation frameworks
3. It's not about replacing developers. It's about amplifying what they can build.
1. Choose your agent patterns carefully
2. Prioritize evaluation frameworks
3. It's not about replacing developers. It's about amplifying what they can build.
5/ And the real-world case studies were incredible:
🎮 Supercell bots handling 100k+ messages/day
✈️ Virgin Atlantic’s voice agents
💼 AI advisors helping investment committees
Agents are already operating at scale.
🎮 Supercell bots handling 100k+ messages/day
✈️ Virgin Atlantic’s voice agents
💼 AI advisors helping investment committees
Agents are already operating at scale.
Agentic AI-MeetUp London #2 Wednesday 15th October, Wed, Oct 15, 2025, 6:00 PM | Meetup
# London Agentic AI Meet-Up #2 - October **Hosted by V7Labs.** **Register now to secure your spot — spaces are limited and will go fast!** Join us on Wednesday, 15th Oct
www.meetup.com
October 1, 2025 at 8:54 AM
5/ And the real-world case studies were incredible:
🎮 Supercell bots handling 100k+ messages/day
✈️ Virgin Atlantic’s voice agents
💼 AI advisors helping investment committees
Agents are already operating at scale.
🎮 Supercell bots handling 100k+ messages/day
✈️ Virgin Atlantic’s voice agents
💼 AI advisors helping investment committees
Agents are already operating at scale.
4/ Julian Kaljuvee from Microsoft went deeper into multi-agent architectures and frameworks like LangGraph. Plus a glimpse at where the ecosystem is heading.
Agentic AI-MeetUp London #2 Wednesday 15th October, Wed, Oct 15, 2025, 6:00 PM | Meetup
# London Agentic AI Meet-Up #2 - October **Hosted by V7Labs.** **Register now to secure your spot — spaces are limited and will go fast!** Join us on Wednesday, 15th Oct
www.meetup.com
October 1, 2025 at 8:54 AM
4/ Julian Kaljuvee from Microsoft went deeper into multi-agent architectures and frameworks like LangGraph. Plus a glimpse at where the ecosystem is heading.
3/ Andrew Liubinas from Tomoro AI broke down a critical distinction:
👉 Router patterns = deterministic logic
👉 Orchestrator patterns = letting LLMs decide
Knowing when to use each is foundational for building reliable agents.
👉 Router patterns = deterministic logic
👉 Orchestrator patterns = letting LLMs decide
Knowing when to use each is foundational for building reliable agents.
Agentic AI-MeetUp London #2 Wednesday 15th October, Wed, Oct 15, 2025, 6:00 PM | Meetup
# London Agentic AI Meet-Up #2 - October **Hosted by V7Labs.** **Register now to secure your spot — spaces are limited and will go fast!** Join us on Wednesday, 15th Oct
www.meetup.com
October 1, 2025 at 8:54 AM
3/ Andrew Liubinas from Tomoro AI broke down a critical distinction:
👉 Router patterns = deterministic logic
👉 Orchestrator patterns = letting LLMs decide
Knowing when to use each is foundational for building reliable agents.
👉 Router patterns = deterministic logic
👉 Orchestrator patterns = letting LLMs decide
Knowing when to use each is foundational for building reliable agents.
2/ We kicked things off with Zilch’s CTO Sean Hederman, showing how to boost developer productivity without sacrificing quality.
Their data science team also revealed how they're reinventing merchant search with semantic AI (not just keywords).
Their data science team also revealed how they're reinventing merchant search with semantic AI (not just keywords).
Agentic AI-MeetUp London #2 Wednesday 15th October, Wed, Oct 15, 2025, 6:00 PM | Meetup
# London Agentic AI Meet-Up #2 - October **Hosted by V7Labs.** **Register now to secure your spot — spaces are limited and will go fast!** Join us on Wednesday, 15th Oct
www.meetup.com
October 1, 2025 at 8:54 AM
2/ We kicked things off with Zilch’s CTO Sean Hederman, showing how to boost developer productivity without sacrificing quality.
Their data science team also revealed how they're reinventing merchant search with semantic AI (not just keywords).
Their data science team also revealed how they're reinventing merchant search with semantic AI (not just keywords).
7/ Full write-up is on the Jentic blog:
👉 jentic.com/blog/do-we-r...
Curious: if you’re running agents in production, how are you evaluating them today?
👉 jentic.com/blog/do-we-r...
Curious: if you’re running agents in production, how are you evaluating them today?
Jentic is building the agentic knowledge layer
One integration, thousands of APIs and workflows. Connect your AI agents to real APIs with secure execution and credential management.
jentic.com
September 19, 2025 at 8:15 AM
7/ Full write-up is on the Jentic blog:
👉 jentic.com/blog/do-we-r...
Curious: if you’re running agents in production, how are you evaluating them today?
👉 jentic.com/blog/do-we-r...
Curious: if you’re running agents in production, how are you evaluating them today?
6/ We also touched on:
Why A/B testing is tricky with non-determinism
How DSPy frames “self-improvement”
Why the “outer loop” matters for agent workflows
Why A/B testing is tricky with non-determinism
How DSPy frames “self-improvement”
Why the “outer loop” matters for agent workflows
Jentic is building the agentic knowledge layer
One integration, thousands of APIs and workflows. Connect your AI agents to real APIs with secure execution and credential management.
jentic.com
September 19, 2025 at 8:15 AM
6/ We also touched on:
Why A/B testing is tricky with non-determinism
How DSPy frames “self-improvement”
Why the “outer loop” matters for agent workflows
Why A/B testing is tricky with non-determinism
How DSPy frames “self-improvement”
Why the “outer loop” matters for agent workflows
5/ The biggest mistake? Overscoping.
Trying to build a universal agent instead of starting narrow.
Narrow use cases are easier to measure, safer to deploy, and deliver faster value.
Trying to build a universal agent instead of starting narrow.
Narrow use cases are easier to measure, safer to deploy, and deliver faster value.
Jentic is building the agentic knowledge layer
One integration, thousands of APIs and workflows. Connect your AI agents to real APIs with secure execution and credential management.
jentic.com
September 19, 2025 at 8:15 AM
5/ The biggest mistake? Overscoping.
Trying to build a universal agent instead of starting narrow.
Narrow use cases are easier to measure, safer to deploy, and deliver faster value.
Trying to build a universal agent instead of starting narrow.
Narrow use cases are easier to measure, safer to deploy, and deliver faster value.
4/ LLM judges can help if calibrated.
Annotate examples yourself
Compare with the judge’s output
Tune until agreement is high
Done right, it accelerates eval loops.
Annotate examples yourself
Compare with the judge’s output
Tune until agreement is high
Done right, it accelerates eval loops.
Jentic is building the agentic knowledge layer
One integration, thousands of APIs and workflows. Connect your AI agents to real APIs with secure execution and credential management.
jentic.com
September 19, 2025 at 8:15 AM
4/ LLM judges can help if calibrated.
Annotate examples yourself
Compare with the judge’s output
Tune until agreement is high
Done right, it accelerates eval loops.
Annotate examples yourself
Compare with the judge’s output
Tune until agreement is high
Done right, it accelerates eval loops.
3/ Enterprises in law, healthcare, and finance can’t skip evals.
Silent failures or infinite loops aren’t just bugs. They’re liabilities.
Silent failures or infinite loops aren’t just bugs. They’re liabilities.
Jentic is building the agentic knowledge layer
One integration, thousands of APIs and workflows. Connect your AI agents to real APIs with secure execution and credential management.
jentic.com
September 19, 2025 at 8:15 AM
3/ Enterprises in law, healthcare, and finance can’t skip evals.
Silent failures or infinite loops aren’t just bugs. They’re liabilities.
Silent failures or infinite loops aren’t just bugs. They’re liabilities.
2/ Startups can start simple:
Collect traces
Label failures (even in Excel)
Run a script to re-test against those examples
That builds the muscle you’ll need later.
Collect traces
Label failures (even in Excel)
Run a script to re-test against those examples
That builds the muscle you’ll need later.
Jentic is building the agentic knowledge layer
One integration, thousands of APIs and workflows. Connect your AI agents to real APIs with secure execution and credential management.
jentic.com
September 19, 2025 at 8:15 AM
2/ Startups can start simple:
Collect traces
Label failures (even in Excel)
Run a script to re-test against those examples
That builds the muscle you’ll need later.
Collect traces
Label failures (even in Excel)
Run a script to re-test against those examples
That builds the muscle you’ll need later.
6/
If you’re curious about:
* AI agents that actually work in production
* APIs and orchestration standards
* Open source projects like the Arazzo Engine
check out the article and tell your thoughts
👉 dev.to/rodriveraai/...
If you’re curious about:
* AI agents that actually work in production
* APIs and orchestration standards
* Open source projects like the Arazzo Engine
check out the article and tell your thoughts
👉 dev.to/rodriveraai/...
Stop Building Brittle Agent Workflows
TL;DR: AI agents fail when they can't reliably chain API calls together. The Arazzo Generator...
dev.to
September 17, 2025 at 8:54 AM
6/
If you’re curious about:
* AI agents that actually work in production
* APIs and orchestration standards
* Open source projects like the Arazzo Engine
check out the article and tell your thoughts
👉 dev.to/rodriveraai/...
If you’re curious about:
* AI agents that actually work in production
* APIs and orchestration standards
* Open source projects like the Arazzo Engine
check out the article and tell your thoughts
👉 dev.to/rodriveraai/...
5/
For me, this is about learning how to show up in new spaces.
I want to see what resonates outside my usual circles and connect with developers where they already hang out.
For me, this is about learning how to show up in new spaces.
I want to see what resonates outside my usual circles and connect with developers where they already hang out.
September 17, 2025 at 8:54 AM
5/
For me, this is about learning how to show up in new spaces.
I want to see what resonates outside my usual circles and connect with developers where they already hang out.
For me, this is about learning how to show up in new spaces.
I want to see what resonates outside my usual circles and connect with developers where they already hang out.
4/
And unlike polished marketing blogs, Dev.to is where people share practical lessons.
They share things they’ve learned building, testing, and shipping.
That’s why it makes sense to bring Arazzo and agent workflows here.
And unlike polished marketing blogs, Dev.to is where people share practical lessons.
They share things they’ve learned building, testing, and shipping.
That’s why it makes sense to bring Arazzo and agent workflows here.
September 17, 2025 at 8:54 AM
4/
And unlike polished marketing blogs, Dev.to is where people share practical lessons.
They share things they’ve learned building, testing, and shipping.
That’s why it makes sense to bring Arazzo and agent workflows here.
And unlike polished marketing blogs, Dev.to is where people share practical lessons.
They share things they’ve learned building, testing, and shipping.
That’s why it makes sense to bring Arazzo and agent workflows here.
3/
Dev.to feels different:
1. It’s community-first, built around sharing and discussion.
2. Posts are accessible to a global audience, from newcomers to senior engineers.
3. Comments matter as much as the article.
Dev.to feels different:
1. It’s community-first, built around sharing and discussion.
2. Posts are accessible to a global audience, from newcomers to senior engineers.
3. Comments matter as much as the article.
September 17, 2025 at 8:54 AM
3/
Dev.to feels different:
1. It’s community-first, built around sharing and discussion.
2. Posts are accessible to a global audience, from newcomers to senior engineers.
3. Comments matter as much as the article.
Dev.to feels different:
1. It’s community-first, built around sharing and discussion.
2. Posts are accessible to a global audience, from newcomers to senior engineers.
3. Comments matter as much as the article.
2/
This is my first time writing for the Dev.to community.
Until now, most of my writing lived on LinkedIn, company blogs, or my own platforms.
Why try something new?
This is my first time writing for the Dev.to community.
Until now, most of my writing lived on LinkedIn, company blogs, or my own platforms.
Why try something new?
September 17, 2025 at 8:54 AM
2/
This is my first time writing for the Dev.to community.
Until now, most of my writing lived on LinkedIn, company blogs, or my own platforms.
Why try something new?
This is my first time writing for the Dev.to community.
Until now, most of my writing lived on LinkedIn, company blogs, or my own platforms.
Why try something new?