An Introduction to Baser Potential
The thesis, the practice, and why the operating layer above your systems of record is where value now compounds.
Talks, essays, and field notes from the practice — plus the outside reading that shapes how we work.
A short overview of the thesis, the practice, and why the operating layer above your systems of record is where value now compounds.
The thesis, the practice, and why the operating layer above your systems of record is where value now compounds.
How many agents are in production, what they are doing, which accountable person each is bound to, whether the company can reconstruct what an agent did, and the exposure if agent systems fail.
Faster coding alone moves the bottleneck to review, security, release, and support. The software factory puts intent, build, and proof in that order.
What a good first candidate looks like, the three shapes most candidates take, examples across six functions, and how to describe one in five lines.
Notes from SVEN’s CXO and AI panel with Tobias Yergin — why agent accountability, visibility and risk are questions every board needs to ask.
A governed path up the AI value ladder. What a McKinsey study of 471 PE-backed companies says about where markets pay for AI — and why governed operations are the way up.
After a $285 billion single-day repricing, which vertical SaaS companies hold their value — the four positions, five plays, and three questions that keep them standing.
A builder's read of Salim Ismail's Organizational Singularity — where we agree, where the essay's paragraph becomes our engineering, and why the morning after is engineering.
Three immediate challenges facing the companies we advise — ungoverned agents, the widening execution gap, and AI-native economics — and a fourth, longer-term threat from the frontier labs.
Software CEOs, boards, and their investors. What the market priced, the three paths it now rewards, the four ways the transition fails, and the capability all three paths share.
Most of what Maryam Ashoori said about AI governance we would sign without edits. Here is the small part where we draw the line differently.
Part two of three. Six moves that carry the discipline of a 34-agent company of one into enterprise infrastructure — policy, authority, evidence, and liability.
Part one of three. What Allie K. Miller's AI workforce actually looks like — an AI chief of staff, six directors, thirty-four agents — and the discipline underneath it.
Different cycles. Different speeds. Uneven absorption. Why intelligence is no longer the binding constraint on what AI delivers — absorption is — and the matching discipline required to bridge the gap.
The most convenient feature your AI vendor ships is the one that quietly becomes your System of Record — and why the Evidence Plane is the only real exit.
On attention, awareness, and the inner posture required to build intelligent systems well. Tobias Yergin reflects on what the Transformer's central insight asks of the humans who deploy it.
A curated set of references — foundations, operating models, governance, and architecture — we return to often.
The thesis that value is migrating from systems of record to a coordinating layer above them.
Argues the next platform shift is a coordinating intelligence layer that sits above today's systems of record.
How Block is restructuring around intelligence as the organizing principle of the company rather than reporting lines.
Original framing of the layer above record/engagement systems and why it becomes the durable moat.
YC's short note on how AI-native startups are being built with radically smaller teams and faster compounding loops.
The foundational essay reframing software as learned weights rather than hand-written code.
Annual industry survey covering research, industry, politics, and safety — the field's de facto yearbook.
Why an organization's loop — observe, orient, decide, act, verify — is the actual unit of competitive advantage.
Boyd's epistemological essay on how analysis and synthesis drive adaptive thinking under uncertainty.
Boyd's distilled briefing on the OODA loop as a contest of orientation and tempo, not mere speed.
How JSOC redesigned itself around shared consciousness and empowered execution to outpace a networked adversary.
How serious teams measure whether an AI system is actually doing the work — and improving.
Anthropic's working approach to evaluating model behavior across capability, safety, and real-world deployment.
The US government's voluntary framework for governing, mapping, measuring, and managing AI risk in production.
The original paper on training models to critique and revise their own outputs against a written set of principles.
Reference material on the orchestration patterns that make agentic systems operable.
Names the emerging discipline that sits between ML research and product engineering — and why it's a distinct role.
A practical taxonomy of evals, RAG, fine-tuning, caching, guardrails, and UX patterns for shipping LLM products.
A clear, end-to-end primer on agent architectures, tool use, planning, and the failure modes that actually matter.
From code-aware evals to self-healing loops — what actually works for evaluating agents in production.