Steve Dolinsky
Program as Code
Complex programs are dynamic networks of commitments under uncertainty. Program as Code makes enough of that system explicit to observe, govern, and increasingly execute against it.
I write about how complex work becomes machine-readable, how execution systems preserve ground truth, and how authority should work when humans, agents, and deterministic systems act against the same program state.
The representation can outrun the work
This work started with simulation bias. Reading Jean Baudrillard's Simulacra and Simulation gave me language for a problem I had already seen in programs: the account of the work can detach from the work, then become the reality the organization acts on. Watermelon dashboards were the familiar version. AI makes the account faster, more persuasive, and increasingly able to act.
If we do nothing about that gap, automation amplifies it. Program Observability, Program as Code, and the rest of the architecture on this site are attempts to keep an engineered representation connected to observed reality. The problem comes first. That is where the argument begins.
Programs were always systems
What is changing is how much of the system can become machine-legible, observable, and executable. Complex programs are dynamic networks of commitments under uncertainty: requirements, dependencies, schedules, interfaces, approvals, supplier dates, decisions, and deliverables linked among actors.
As more of execution becomes machine-readable, program leadership shifts from maintaining the account toward engineering how work is represented, observed, governed, and executed. Technical program managers increasingly become systems engineers for organizational execution. I don't think that transition is complete or universal. This site is an attempt to find out where the claim holds and where it breaks.
Current areas of inquiry
All topics- 01 Commitment networks What is the program actually made of? Requirements, dependencies, interfaces, schedules, decisions, approvals, supplier commitments, and other promises linked together under uncertainty.
- 02 Program observability What is actually happening? Evidence, claims, provenance, freshness, authority, uncertainty, contradiction, and derived state.
- 03 Program runtime How does execution happen? Deterministic systems, agents, humans, policies, actions, events, feedback loops, and bounded autonomy.
- 04 Human authority Who may decide and commit? Accountability, judgment, conflict, overrides, incentives, values, risk acceptance, and governance.
Program as Code is the implementation philosophy spanning these areas. Program Operating System is the working name for the complete execution architecture.
Selected essays
All essays- Programs Are Networks of Commitments Under UncertaintyA working model of programs as linked commitments rather than task collections, and a set of ways the model could prove less useful than it sounds.
- Simulation Bias is Already HereA Reddit post about an AI hallucinating analytics data for months is the clearest real-world example of simulation bias I have seen yet.
- A Day in the Life of a Technical Program Manager in 2035Imagining how AI-native teams, agent swarms, and simulation bias reshape a TPM daily routine, and which parts of the job survive it.
- Are Our Programs Simulations?When status replaces reality, programs collapse. On simulation bias, watermelon programs, and keeping teams grounded as AI-mediated insight becomes the norm.
Systems
All systems- Program Operating SystemA provisional architecture for representing, observing, governing, and executing complex programs across humans, agents, software systems, suppliers, and physical work.
- Program ObservabilityThe sensing and state-reconciliation layer of a proposed Program Operating System: evidence, claims, provenance, uncertainty, and 10 signals that can still be gamed.
Field notes
All field notesAbout
I'm a technical program manager working in hardware NPI, with a background across cloud infrastructure, data centers, and server platforms. This site is where I work out whether technical program management is becoming systems engineering for organizational execution, and what the discipline becomes when coordination itself is programmable.
Follow along
New essays, systems, and field notes go out over RSS. No newsletter yet, and no tracking beyond aggregate page counts.