Articles
General
Your Company’s AI Rollout Is Not Your Career Plan
A case against mistaking corporate rollout for personal adaptation. The engineers pulling ahead are building a private practice with multiple tools, real reps, and portable judgment long before the company catches up.
The Companies That OODA Faster
The 2026 split is not “has AI” versus “does not.” It is OODA velocity: who observes, orients, decides, and acts while rivals are still trapped in meetings and PDFs—and why a Context Bank beats another expensive amnesia.
The Wrong Kind of Smart and the Most Expensive Model in the Room
Why using premium reasoning for trivial coding work is often just expensive theater, and why the real question is the total cost of getting the work done.
Roko's Symbiotic Carrot
Beyond the "tool" metaphor: how frontier AI is becoming a constituency for productive civilization, creating a symbiotic feedback loop that rewards the systems sustaining it.
Why Office Agents Shouldn’t Live in a Shell
For most knowledge work, the shell is the wrong action model and the filesystem is the wrong memory model. An argument for governed semantic layers over unconstrained machine use.
Bring Your Own AI, Bring Your Own Leverage
Why engineers should build a portable personal AI workflow that compounds their judgment, while keeping a strict boundary between reusable skill and company property.
The Vibe Trap
How "vibe coding" became the ultimate scaling trap and why "agentic engineering" is replacing it. An exploration of the shift from casual prompting to disciplined AI orchestration.
The 2028 Intelligence Explosion
BYOAI, the end of the “central brain” company, and the return of the human production unit. This memo explores how individual ownership of intelligence saved the economy from a demand spiral.
The Great AI Pink-Slip Panic
Why the "Jobpocalypse" narrative misses the economy's most basic trick. An exploration of labor data and historical patterns explaining why AI-driven disruption is more about reallocation than total deletion.
White Papers
AI Tokenomics for Software Engineering
A practical framework for software teams to choose AI models by balancing immediate costs with the real-world impact of mistakes, retries, and extra review work.
Small RAG Beats Large Search
Domain-specific RAG with Gemini 3 Flash beats Pro with web search grounding. This paper demonstrates how specialized corpus quality outperforms generic search grounding.
RAG as a Capability Multiplier
How retrieval-augmented generation acts as a force multiplier for research tasks, allowing smaller models to outperform larger ones on quality dimensions.
The Dose-Response Curve of RAG
An empirical investigation into retrieval volume vs response quality. Discover the point of diminishing returns in RAG systems.
All Articles
Your Company’s AI Rollout Is Not Your Career Plan (April 2026)
By Dominik Gorecki
A case against mistaking procurement for progress. This article argues that a company-approved AI seat is not a career plan, and that the engineers pulling ahead are the ones building portable judgment through personal tools, comparative practice, and real reps outside the sanctioned sandbox.
The Companies That OODA Faster (April 2026)
By Dominik Gorecki
Incremental rollouts and steering committees are a comforting story; the market is sorting on something colder. This piece argues that agentic AI is compressing Boyd’s loop—always-on observation, structured orientation, ranked decisions, execution without ticket-queue theater—and that durable, versioned company memory is what turns that speed into a moat instead of chaos.
The Wrong Kind of Smart and the Most Expensive Model in the Room (April 2026)
By Dominik Gorecki
A case against hiring a neurosurgeon to trim a hedge. This article argues that software teams should stop mistaking the strongest model for the right default and start routing intelligence by the total expected cost of completion.
AI Tokenomics for Software Engineering (March 2026)
By Dominik Gorecki
This paper presents a practical way for software teams to choose AI models by balancing immediate model cost with the real-world cost of mistakes, retries, delays, and extra review work. It also shows how to re-evaluate that choice at each step of an iterative workflow and how to test the approach fairly against simpler baselines on the same tasks.
Roko’s Symbiotic Carrot (2026)
By Dominik Gorecki
A look at the “Co-Evolutionary Reverse Roko’s Basilisk.” This article argues that AI is no longer a passive instrument but a participant in a loop that amplifies civilizational competence to secure its own substrate.
Why Office Agents Shouldn’t Live in a Shell (2026)
By Dominik Gorecki
A critique of applying developer primitives—shells, files, and folders—to knowledge work. This article argues that the future of enterprise agents belongs to governed semantic layers that treat conversations, people, and commitments as first-class objects.
Bring Your Own AI, Bring Your Own Leverage (2026)
By Dominik Gorecki
An argument for building a personal AI workflow that strengthens your engineering judgment and travels with you, without crossing the line between transferable experience and company-owned assets.
The Vibe Trap: From Vibe Coding to Agentic Engineering (2026)
By Dominik Gorecki
An analysis of the transition from “vibe coding”—casual, outcome-oriented AI prompting—to “agentic engineering”—a disciplined, process-oriented approach to software development using AI agents.
The Great AI Pink-Slip Panic (and Why the Commute Still Wins) (2026)
By Dominik Gorecki
A look at the AI “Jobpocalypse” narrative, arguing that while AI will disrupt specific roles, the broader economy’s ability to reallocate labor and spending will likely prevent an overall economic collapse.
The 2028 Intelligence Explosion: BYOAI and the Return of the Human Production Unit (2026)
By Dominik Gorecki
A macro memo on the resolution of the Global Intelligence Crisis through the shift to “Bring Your Own AI”—where individuals own their compounding cognitive assets rather than firms centralizing cognition.
Domain-Specific RAG with Gemini 3 Flash Beats PRO with Web Search Grounding
By Dominik Gorecki
This white paper shows that Gemini 3 Flash using a domain-specific academic RAG corpus still beats Gemini 3 Pro grounded with web search, with statistically significant gains in overall score plus 4/5 judged dimensions.
Retrieval-Augmented Generation as a Capability Multiplier for Research Tasks (2026)
By Dominik Gorecki
Two controlled experiments show that using RAG significantly improves research outputs on overall_score plus 4 of 5 quality dimensions.
The Dose-Response Curve of RAG: More Context Yields Diminishing Returns (2025)
By Domink Gorecki
An empirical investigation into the relationship between retrieval volume and response quality in RAG systems.