Insights
Notes on building systems that hold up.
Short, practical writing from Hermida Technologies on engineering, automation, AI in operations, security, and the client experience of working software.
Latest writing
Recent articles.
- Security6 min read
Autonomous AI Gone Wild: Rogue Agent Behavior Sparks Urgent Enterprise Governance Demands
A wave of unexpected agent failures—from unauthorized invoice generation to aggressive scraping—pushes EU regulators and cybersecurity teams to enforce real-time agent guardrails.
Read article - Engineering5 min read
Arm Unveils Neoverse N4: 128 Cores on TSMC N3P to Break the AI Inference Compute Wall
Arm has pulled back the curtain on its Neoverse CSS N4 platform, packing 128 cores on TSMC's N3P node with PCIe 7.0 and CXL 4.0 to tackle soaring agentic AI inference demands.
Read article - AI & operations5 min read
On-Device SLMs Reach Parity: Why MiniCPM5-2B Signals the End of Cloud-First Dominance
OpenBMB's open-source MiniCPM5-2B outscores older 4B models while running locally on devices, marking a turning point in the economics of enterprise AI workloads.
Read article - Security6 min read
Securing the Agentic Perimeter: Mitigating Memory Poisoning and Goal Hijacking in 2026
With autonomous enterprise agents managing production workflows, cybersecurity teams must counter sophisticated threats targeting agent memory pools and multi-agent communication networks.
Read article - Engineering6 min read
Beyond Vibe Coding: Why Engineering Teams Are Adopting Objective-Validation Protocols
As autonomous AI agents replace casual coding assistants, enterprise engineering organizations are abandoning vibe coding in favor of structured Objective-Validation Protocols.
Read article - Tech news5 min read
The John Ternus Era at Apple: Navigating Hardware Supremacy Without a Frontier AI Model
As John Ternus officially assumes the role of Apple chief executive, the tech giant faces an urgent architectural test: sustaining premium ecosystem margins without an in-house frontier model.
Read article - Engineering5 min read
HPE and Oracle’s AI Alliance Exposes the Reality of Hybrid Enterprise Infrastructure
HPE's strategic tie-up with Oracle highlights why enterprise AI cannot live entirely in the public cloud, prompting IT teams to re-evaluate on-premises data gravity.
Read article - AI & operations6 min read
The Death of the SaaS Seat: How Multi-Agent Systems Are Upending Enterprise Software Pricing
New Silicon Valley Bank data reveals a rapid exodus from per-user seat pricing toward hybrid and outcome-based models as autonomous AI agents replace human software seats.
Read article - Security5 min read
Anthropic Solves the Enterprise AI Deadlock: Why Customer-Controlled Storage Changes the Game
Anthropic's newly launched Enterprise File System architecture decouples model compute from proprietary data storage, providing a breakthrough blueprint for zero-data-retention compliance.
Read article - AI & operations5 min read
The Post-Detection Dilemma: Why AI Watermarking Is Failing Enterprise Trust
Pangram's high-profile deployments spotlight a harsh reality: simple binary AI detection is obsolete. As automated text and images proliferate, trust requires structural provenance.
Read article - Tech news5 min read
The John Ternus Era at Apple Begins: Hardware Realism Meets the AI Frontier Deficit
As John Ternus assumes Apple's CEO seat, Cupertino confronts an unprecedented challenge: orchestrating an AI transformation without owning an in-house frontier foundation model.
Read article - Engineering6 min read
The $31 Trillion Data Center Reality Check: Orchestration Fabrics Confront the AI Debt Bomb
With long-term AI data center expenditures projected to hit $31.6 trillion, enterprise infrastructure is hitting a physical and orchestration wall. Here is why the switchboard layer matters.
Read article - Client experience5 min read
The Desktop War Moves to Slack: Inside Adobe’s Radical MCP-Powered Platform Bet
Adobe is embedding over 70 creative and productivity applications directly into Slack via the Model Context Protocol. The move signals a broader shift away from monolithic desktop apps toward headless conversational tool distribution.
Read article - Tech news6 min read
The Power Reality: Big Tech’s AI Ambitions Run Headlong into the Nuclear Clock
With AI inference hardware demanding unprecedented megawatts, hyperscalers are racing to sign nuclear power purchase agreements. But as small modular reactors face regulatory lags, the compute industry confronts an inescapable physics problem.
Read article - AI & operations5 min read
The Agent Control Plane War: Boomi and Broadcom Move to Tame Enterprise Agent Sprawl
As autonomous agents flood corporate environments, tech giants are pivoting from model development to control planes. The real enterprise AI battle is no longer raw intelligence—it is governance, runtime containment, and context engineering.
Read article - Security6 min read
Defense Deployments and Digital Provenance: The Enterprise Imperative for Agentic Security
With government agencies rolling out custom LLMs and detection startups pulling fresh capital, securing autonomous agents and authenticating content provenance have become critical enterprise mandates.
Read article - AI & operations6 min read
Anthropic's $80 Billion Infrastructure Sprint: The Industrialization of Enterprise AI Workloads
Anthropic's massive multi-billion-dollar compute commitments signal a decisive evolution from speculative model training to high-throughput, private enterprise compute infrastructure.
Read article - Tech news5 min read
Brussels Designates ChatGPT as a Very Large Search Engine: What the DSA Reclassification Means for Enterprise AI
The European Commission's reclassification of ChatGPT as a Very Large Online Search Engine introduces unprecedented algorithmic auditing, setting off regulatory ripple effects for enterprise deployments.
Read article - AI & operations5 min read
The Agent Deployment Bottleneck: Why CI/CD Pipelines Are Becoming the New Frontier for Enterprise AI
Despite massive investments in autonomous agents, enterprise adoption remains trapped in development purgatory. A shift toward agent-native CI/CD tooling promises to bridge the gap.
Read article - Automation5 min read
Before you automate, map the system
Most automation failures are not tool failures. They happen when a process gets automated before anyone understands what the process actually is.
Read article - AI & operations6 min read
Where AI actually earns its place in operations
AI is genuinely useful in operations — in specific places, with specific guardrails. Here is where it pays off and where it quietly creates risk.
Read article - Engineering5 min read
Build the smallest system that solves the problem
The best system is the smallest one that fully solves the problem — small enough to understand, small enough to maintain, small enough to trust.
Read article - Security4 min read
Access control belongs in the first commit
Retrofitting security never works as well as building it in. Access decisions belong server-side, in the first version of every system.
Read article - Client experience4 min read
A client portal is a promise, not a feature
A client portal sets a standing expectation: that the client can see the truth of the engagement at any moment. Build one only if the operation can keep that promise.
Read article
Want this thinking applied to your operation?
The articles describe how we work. A conversation is how it starts.