and we are back to MCP!
It shouldn’t surprise anyone at this point in time that anything AI related had so many ups and downs
“Let’s pace AI!” - “We need slow down!” - “AI will kill all of us”
Sounds a lot like we can’t sustain the expenditure on compute (with energy implications)
Wondering how open source models (China?) will react.
After a little over 30 years building software in different contexts for different domains and audiences…
It has been a really interesting and FUN journey to re-think my practice and figure out how to build software for agents.
How are you managing social media these days?
I’d love to be here more, like the “old times.”
But between creating, reading, engaging here, keeping up with the more “commercial” @ LinkedIn and, well… actually working, it’s hard.
And delegating it to AI defeats the point.
Finally got access to Astra…
Not long after: boom! 💥
Claude just reset!
Gotta love these Labs fights 😁
What a time to be a builder!
Thinking… (reading the comments around the Java documentary)
From where I sit:
JBoss ended up at IBM.
Red Hat ended up at IBM.
Spring ended up at Broadcom.
GitHub ended up at Microsoft.
But sure… open source won. 🤷♂️
I’m genuinely impressed by OpenAI’s progress. A few months ago, it was hard to imagine a real alternative to Claude Code. Today, Codex is much better, and the models are excellent.
I just wish working with OpenAI were easier. Support is nearly impossible, and the business offerings are underwhelming. (I've been `forced` to use a personal account)
Still, I’m hopeful: if they can transform the product and models this much, they can transform the customer experience too.
Happy 250th Anniversary, America 🇺🇸🎉
This is my first 4th of July as an American citizen, after more than a decade calling this country home.
Grateful for the life, family, friendships, and opportunities we’ve built here, and proud to keep contributing in my own way.
early this week I announced the Apache KIE 10.2.0 release
Apache KIE is the umbrella project for Drools, jBPM, Kogito and OptaPlanner
kie.apache.org/blog/kie_10_2…
and yesterday I also announced the release of @aletyxai Kogito AI AddOns that enables Kogito workflows to coordinate and orchestrate AI Agents alongside humans.
github.com/aletyx/aletyx-kog…
Demo from my presentation of @arcofai that leveraged the Kogito AI AddOn:
github.com/porcelli/arc-of-a…
"The model said so" isn't an answer regulators and auditors are looking for.
That's why AI is still locked out of the high-stakes core of the business. Loan approvals. Insurance claims. Regulated workflows. Healthcare.
I'll be talking about how to change that at @QCon Boston AI on June 2.
Live on stage: a corporate decision model driving Agents, Skills, and Guardrails with NVIDIA NeMo. That's how the agentic stack becomes accountable.
boston.qcon.ai/speakers/alex…
it’s a wrap!
this 2nd edition consolidated @arcofai as a must-attend conference on the circuit.
grateful to have presented at both editions, and already looking forward to coming back next year with more to share on AI.
#ArcOfAI 2026 was one for the books! 🤖
An amazing week packed with learning, big ideas, and an incredible community. From deep-dive sessions to hallway moments—so much to take away.
Huge thanks to our speakers, attendees, & everyone behind the scenes.
🚀 See you next year!
Yesterday I had the privilege to present at @arcofai about bringing agents into mission-critical systems in a governed way.
Demo video coming this week.
slides: aletyx.link/Arc-of-AI-Alex-P…
Three patterns I explored:
1. Cluster tasks that agents handle. When the agent can't, a human steps in.
2. Add a deterministic gate after those clusters.
3. When the gate conflicts with the AI recommendation, bring a human into the loop with full context.











