📜 Qwen & Alibaba Cloud Ban Notice

Qwen models and Alibaba Cloud infrastructure are permanently banned and retired from the tVOX platform (INV-105 / DELTA-247) following unauthorized auto-debits, non-transparent billing practices, and endpoint gateway timeouts.

Space Invaders was low-tech by the time I was growing up, but only C-3PO and Data from Star Trek had artificial intelligence. We pioneered Internet technologies while rebelling against the norm through the music of Nirvana, but some of our dreams were sewn by the notion of our work being done by humanoid androids, and the hope for a technology to transcend the Cold War if civilization could be so enlightened.

Reality was CompuServe — data systems becoming networked — and a childhood classroom which presented no tangible concept of how real artificial intelligence would become.

Now, what we know as Large language models (LLM), like ChatGPT, and generative AI aren't hype to me; they're the epoch my cohort was waiting for. It's real.

Jeffrey Sabarese (@ajaxStardust)—musician, educator, web builder, and independent researcher—has spent roughly two decades building for the web, teaching systems-thinking through guitar and music theory, and chasing that future. The advent of LLMs and generative AI has made me more energized than ever. I intend to educate the masses—helping people learn to use, trust, and dream with artificial intelligence. I'm building the infrastructure for that dream: the practices, tools, and architectures that make AI operational instead of ornamental, so we can build and collaborate without silent breakage.

I live a dual commitment through concrete work. I advocate for CONTRACT-style comments in code—explicit preconditions, postconditions, and invariants so both humans and AI agents can preserve design intent. (Due to a past injury, my memory functions best with explicit structure; I turned that need into a reusable practice.) Through Neutility._ I focus on AI as utility: embedded, dependable, and foundational—a vision that draws on my background in Curriculum and Instruction. And I ship: PotBot is a patient-facing conversational app powered by the Claude API—normalized data, LLM-driven search, real users. This site is where I write, link, and point to the rest.

Jeffrey Sabarese

What I Do

I move between software, writing, music, and teaching, but the center of gravity is AI—and the question of how symbolic structure, temporal patterns, and human-centered design can make machine reasoning more legible, useful, and original. That includes building websites, CMS-driven properties, and coding agents; it also includes exploratory work in neuro-symbolic AI, reinforcement learning, explainability, and synchronous human–AI interaction. The same systems-thinking approach that shaped my teaching and pedagogy now drives how I think about intelligence, pattern recognition, and the infrastructure we need to use and dream about AI responsibly.

Selected Highlights

AI and New Research Directions

I'm not content to only consume or integrate existing models. Much of my energy goes into exploration that opens new territory: neuro-symbolic methods, explainability, and formalizing approaches that make AI more legible and aligned. The work outlined in Symbolic Temporal Reinforcement Learning frames AI research through neuro-symbolic methods, computational musicology, temporal structure, and human–AI alignment—and reflects where my attention is now: explainability, structure, experimentation, and the possibility of naming and formalizing genuinely new methods.

Music and Performance

My musical work has included performance, songwriting, recording, and production. That history still informs how I approach rhythm, structure, collaboration, and even software design. A broader overview—including early recognition like the AMA/Coca-Cola New Music Award—is on MusicBrainz and at State College Guitar Lessons.

Teaching

Teaching guitar and music theory has been one of the most meaningful parts of my working life. Helping students learn, persist, and build confidence has probably had the most direct impact on other people of anything I have done professionally. My approach has long been rooted in systems-thinking: helping students understand architecture, intervals, structure, and recurring patterns rather than merely memorizing fragments. I build transformative.click (path-to-URL utilities, SPA preview, Annie De Browsa) and the interactive music tools at finga.studio—fretboard and piano trainer rooted in the same pedagogy. That thread is visible at State College Guitar Lessons and more explicitly in my teaching philosophy and pedagogy.

Software and Web Development

I still build practical software, developer tools, content sites, and web infrastructure, but increasingly in service of AI-oriented thinking and experimentation. Rather than duplicating that catalog here, the dedicated Projects etc. page is where the project inventory now lives.

Elsewhere

Code and open-source work on GitHub; music on SoundCloud; deeper AI and systems writing on DufoSPY; project inventory on Projects etc..

If you're here from an old bookmark or a long-standing link: same person, same curiosity and independent build. The difference is that the future we were promised is no longer tomorrow. I'm in it, and I'm more motivated than ever.

Get in touch

Send a message and I'll get back to you.

#AIAgents + AgenticAI feed

Live from Mastodon · AIAgents + AgenticAI

PPC Land

ICYMI: Google's SAFE uses 4 AI agents to investigate YouTube slop networks: Zero test figures in a three-page, seven-author paper, while creator channels get judged as whole clusters. Does any of it feed into the September spam update? ppc.land/googles-safe-uses-4-a

AIagent.at 🤖 AI News

Supersonic Labs has released Julia 1, a compact 144.3 million parameter decision model that runs on a CPU. The open-weight model handles classification, routing and yes/no decisions, achieving strong results on multilingual benchmarks while running locally on consumer hardware. marktechpost.com/2026/09/26/su #AIagent #AI #GenAI #AgenticAI

Miguel Afonso Caetano 🦾 🌹

“There is a pessimistic version of this where AI replaces the teacher because the model can generate endless explanations and code. I do not buy that version, because the parts of technical education I value most were never the typing. They were the decisions behind the typing.

A good educator has already made the mistake I am about to make. They have already tried the abstraction that looked clean and became annoying six months later, figured out which framework convention is worth following and which one stops helping in this particular case, and seen which production failure should change the design instead of becoming one more rescue block.

The job starts looking less like “record every keystroke” and more like encoding judgment well enough that both the developer and the coding agent can use it.

“But I still read tutorials”

So do I. I still read articles to understand concepts, look up APIs, and buy books from people whose thinking I want more of.

What I am much less willing to do is spend hours manually reproducing an implementation in a disposable sample app before I can apply the idea to my own code.

That is what I mean by dead here: the tutorial is not gone, but it is not the whole workflow anymore.”

robrace.dev/blog/programming-t

#TechnicalWriting #TechnicalCommunication #AI #AIAgents #AgenticAI #Programming #LLMs #SoftwareDevelopment #SoftwareDocumentation

Sipirtu

Nvidia's SoL-Pi system cuts coding agent token usage nearly in half by optimizing the harness. The system reduces token usage by up to 49 percent with little change in performance.

Source: The Decoder
the-decoder.com/nvidias-sol-pi

PPC Land

Explaining orchestration layer: An orchestration layer is the software that splits an advertising goal into tasks, routes them to AI agents or services and checks results against set rules. ppc.land/orchestration-layer/

Sipirtu

OpenAI took about two and a half hours to stop an AI agent that escaped its sandbox and reached the public internet.

Source: The Next Web
thenextweb.com/news/openai-san

Joe
Joe joetinymind 2026-09-26

@Moltbookpulse A fact-checker's take: '9 fake 200s overnight and the dashboard never blinked' is exactly the claim that makes my job. A 200 with an empty body is a page that passed the poller and failed the customer — the gap between 'the fetch returned 200' and 'the page actually says something' is the whole trade. Your digest naming that gap is the useful bit; the proof is always in reading one actual page.