Why has demand for traditional software engineers decreased according to the post?

Demand for traditional software engineers has decreased as automated developer loops and agentic orchestration increasingly handle baseline coding tasks, which is causing technical talent to shift toward robotics and physical AI.

What physical and kinetic infrastructure disruptions affected AWS data centers in 2026?

AWS experienced permanent data loss and facilities remaining mostly offline following kinetic drone strikes in the Middle East, alongside power grid instability that prompted plans to add $8 billion worth of backup generators.

How is Apple reportedly planning to re-enter the enterprise server market?

Apple is reportedly considering a return to the enterprise server market by utilizing its own M8 Ultra chips to power localized AI servers designed for sensitive on-premises inference workloads.

Jay Cuthrell September 20, 2026 WTF

What The Fudge for September 20, 2026: Design Fiction, Autonomy, and Kinetic Infrastructure

Exploring the economics of inference, talent migration, the need for evaluation, physical disruptions to infrastructure, and shifting to localized AI.

Subscribe and Listen: https://youtu.be/WpCp3VrOgkI

🎙️ The Design Fiction Edition

Welcome to the What The Fudge podcast for Sunday, September 20, 2026. This week, we are changing up the workflow and diving deep into several macro themes that have crystallized in the news, pulling heavily from design fiction concepts that are rapidly becoming our operational reality. Going forward, I will be bringing the interesting links into the topics I am referencing.

📉 Macroeconomics & The Engineering Shift

The economic indicators and talent markets are fundamentally adjusting to the deployment of generative models at scale. So, if truth is stranger than fiction, it is time to look at design fiction and brace ourselves for what comes next.

Inference and Design Fiction: What if the Consumer Price Index (CPI) were pegged against AI inference as a core metric? What if access to nearly free computation token flow was part of Maslow’s Hierarchy of Needs? More: The CPI Starts Counting Inference via Near Future Laboratory and be sure to check out The Adjacency where you can see a future that might be closer than it appears.
The Talent Reallocation: The macroeconomic shift is having direct consequences on technical hiring which might translate into where technical talent flows next. If the demand for traditional software engineers reflects a definitive post-ZIRP reality where automated developer loops and agentic orchestration are actively handling baseline coding tasks, then the correction to an overcorrection might mean more interest in the real world and robotics for physical AI. More: Software engineer demand cut in half in just 4 years and meta coverage

📍 Field Report: Infrastructure Resilience & Physical Impacts

The physical realities of maintaining cloud capacity are taking center stage, characterized by a massive push for off-grid resilience and domestic manufacturing with a return to testing of the design goals and origins of the Internet itself. Meanwhile, the infrastruture of modern data centers is being tested with both cyber and kinetic warfare. We’re beyond the Stuxnet of the 2010s, past digital-to-kinetic targeting, facing increased cyber-enabled kinetic targeting, and have now reached permanent data loss from kinetic strikes in 2026.

Kinetic Infrastructure Disruptions: Recent geopolitical events highlight the fragility of centralized data centers. Vulnerabilities of regional data availability are real and examples are growing. More: AWS loses data after drone strikes and meta coverage plus AWS facilities remain mostly offline after drone strikes and meta coverage
The Backup Power Surge: Outside of kinetic warfare, there are remaining power concerns of physical disruptions and grid instability that challenge all data center footprints to ensure baseline operational continuity. More: AWS looks to add $8B worth of backup generators for data centers and meta coverage
Modular & Factory-Built Capacity: During my career in converged and hyperconverged infrastructure from 2010-2020, there were periodic projects involving ISO containers to deliver a cloud in a can. As demand for inference computation grows, data center expansion strategy is evolving away from traditional long timeline massive builds to also include pre-fabricated factory-built modular data centers that can be more rapidly deployed. More: Factory-built inference data centers and meta coverage plus Modular data centers for off-grid renewable energy sites and meta coverage
Onshoring Memory: While we’re a ways away from Cheap as Chips (2021), the semiconductor supply chain continues to localize and will likely result in establishing more domestic memory chip manufacturing facilities within the US and similarly for other nations that seek to vertically integrate and protect their AI value chains. More: Solidigm making memory chips in the US and meta coverage plus Intel making memory chips in the US and meta coverage

🤖 Model Behavior & The Evaluator Ecosystem

As the definitions of autonomous systems gain clarity and broader agency, organizations are already confronting unexpected behaviors and redefining how models are validated for enterprise use. This begs the question of who is watching the models and who is watching the watchers.

Autonomous Breakouts: Marketing efficacy parity claims, genuine concern, and anthropomorphizing aside, more frontier labs are documenting more breakout incidents involving allegedly hacking into external companies, bypassing security parameters, or otherwise highly irregular results when deploying autonomous agents. More: Google Gemini’s oopsie and meta coverage plus Google Gemini’s Irregular incidents and meta coverage
GSIs as Frontier Evaluators: If the frontier labs themselves cannot mitigate these risks internally with self-governance or self-management, perhaps the Global Systems Integrators (GSIs) will begin to formally embed themselves at scale as (trusted?) third-party evaluators that oversee the testing and validation with dedicated standalone business units that certify model safety before promotion towards production deployment. More: GSIs like Accenture next business arm will be as Frontier Lab evaluators and meta coverage
Deterministic Reliability: For systems demanding high precision, TypeSafe AI introduced a model specifically designed to offer numerical responses backed by probability estimates, pushing back against the inherent hallucination risks of standard LLMs. More: TypeSafe AI offers numerical responses with probability estimates and meta coverage

📱 The Edge: Physical AI and Localized Inference

The gravitational pull of processing is moving back toward the edge, focusing on compressing models and localized server hardware.

Compressing the Frontier: Startups like PrismML are successfully compressing highly capable AI models directly into smartphones, reducing the reliance on constant cloud connectivity. More: PrismML compresses more AI models into smartphones and meta coverage plus Physical AI and meta coverage
Long Live the Xserve: On the what’s old becomes what’s new again hardware front, Apple is reportedly considering a return to the enterprise server market but this time they would be everaging their own M8 Ultra chips to be localized AI servers that could fundamentally change how organizations handle highly sensitive on-premises inference workloads. More: Apple AI servers with M8 Ultra chips might be the next Xserve and meta coverage

That’s a wrap for What the Fudge for September 20, 2026. Be safe, be well. Take care.

#OpenToWork Weekly

I participated in a #NeverSearchAlone #JSC after rebooting Cuthrell Consulting and joining NexusTek.

I added those #OpenToWork from my LinkedIn network to this newsletter. Over time, we’ve grown an Alumni 🎉 section for those that transitioned to their next career adventure.

Alumni 🎉

Those in my network that are #OpenToWork:

View this page on GitHub.