The first app in my account was a mood tracker I labeled, literally, "my first app." Now I operate a self-generating education platform, a commercial product built on it, and the infrastructure under both. This is both halves of that change: the craft, and the judgment that directs it.
David George BraunFounder & Executive Engineer, Eudai Gestalt Integrationsworkflowtech.ai · Medellín, CO
7moDec 2025 to Jul 2026
16repositories, one account
1,761commits across the portfolio
580merged pull requests
4systems live in production
8+languages, Python to PowerShell
Certified by AnthropicFormal recognition across the Claude Code partner, Claude Certified Architect, and Claude Certified Developer tracks, the same practice this paper documents.
The story these repositories tell is a ramp. It starts with small experiments, each one teaching a single lesson. A mood registry. A registry backup script. A media stack. A capability tracker. A relationship app. None of them was the point. Each was a rep.
By spring the ambition consolidated into a stack. I started the chatbot framework, then the education engine on top of it, then the commercial fitness module inside that, then the operator toolkit I use to build all of it. In May and June alone I founded seven of the sixteen repositories, in rapid succession, because by then I was building platforms instead of pages.
A second thread ran in parallel. The Anthropic path started with the AI Fluency course and its 4D framework: Delegation, Description, Discernment, and Diligence. Those four became the working discipline behind the build, and the road they opened ran through Anthropic Academy to Claude Certified Architect and a Partner badge.
2025 · run-up
Emotional_Event_Registry · "my first app"regbakROFLCopterCinebrauncapaz
Dec 2025
closer · intimacy generatorbenesseremedestetica · wellness web
Jan 2026
tcba · site & almanac
Apr 2026
EGI-Codex · automation catalog
May 2026
EGI_bot · workflowtech.aiGestaltWorkframe · the frameworkGestaltWorkframeEDU · the platform
GestaltWorkframeEDU · commits per monththe depth ramp
14
May
249
Jun
490
Jul*
753 commits and 366 merged pull requests on the flagship. *July measured through the 17th, already double June's full-month total.
02 · The Craft
Seven technical capability areas, each stated as a claim and backed by what actually shipped. The judgment that directs them comes next.
Capability
LLM-native systems, grounded in real sources
I learned to treat generation as the product, not a feature bolted onto an app, and to hold it to a standard of evidence.
Built a self-generating curriculum engine that takes a subject, infers the pedagogy and domain shape, and produces graded programs with assessments.
Grounded every generation in a real citation corpus, proven on production: 28MB of real book text across 80 sources, cited, with zero fabricated identifiers.
Replaced a single generalist generator with a concurrent expert swarm, one corpus-scoped specialist per discipline: strength, mobility, conditioning, self-defense, yoga, breath, recovery.
Made background generation survive deploys, re-dispatching orphaned runs on restart instead of leaving them stuck.
Hardest lessonA generation that reports success can still be wrong. I verify by walking the finished program through real endpoints, which caught crashes that a 96 percent progress bar never would.
Capability
Full-stack product engineering
I ship the whole stack, database to browser, on one consistent modern toolchain.
Intake, session logging, adaptive evaluation, insights, and locale resolution, all shipped as reusable engine services.
ReachThis same stack recurs across the account: the learning platform, the wellness site, the relationship app, and the capability tracker.
Capability
Commercial & monetization engineering
I built the money layer, with real payment integrations and real unit economics.
Freemium architecture: a free multi-week teaser, the full program behind a paywall.
An entitlement spine tracking status, source, and period-end per user and module.
Stripe checkout across four billing durations with webhook reconciliation, plus PayPal with real webhook signature verification.
A regeneration credit ledger and reuse-first economics: catalog reuse makes zero extra model calls, cutting the cost of a full program generation from $3.40 to under a dollar.
Principle earnedThe cheapest model call is the one you avoid. Reuse before you regenerate.
Capability
Platform & module architecture
I built a platform that builds programs, then made programs snap into it cleanly.
A domain-agnostic engine with content modules that carry all domain config and no engine code.
A module toolchain: .gwmodule bundles, a modulectl CLI, a framework compatibility gate, and atomic install.
Surface separation enforced in code so the platform host and a program host never leak into each other.
Framework reuse by code rather than dependency, keeping the upstream framework and the education engine independently shippable.
Capability
Infrastructure & DevOps
I run the metal the products live on, and I keep it running.
A private WireGuard mesh linking the workstation, two IONOS VPS hosts, and a NAS.
Self-hosted GitHub Actions runners per private repo, parallel pytest with xdist, path-filtered workflows, concurrency auto-cancel.
systemd services, nginx with per-endpoint routing and certbot TLS, Docker images, and a manual gated production deploy.
Operations discipline earned the hard way: reconciling SQLite schema drift on boot, tuning VPS swap after an out-of-memory hang, and diagnosing a runner that reads offline while its service shows active.
Capability
Engineering process & verification
Volume with discipline. Nearly 500 merged pull requests, each through the same gate.
Branch, PR, CI green, coverage gate at 80 percent, squash-merge. Clean history as a standing rule.
Verify against reality: run it before claiming it works.
Diagnose before fixing: gather state, logs, and config first, and keep ruled-out paths closed.
Hunt silent failures: a green PR can still take production down, so I check the checks.
Capability
Range, and the tools I build with
I built the operating system I use to build everything else.
llm-builder-kit: my own doctrine, Next.js templates, and a token-gated operator control plane for building serious software solo, local-LLM-first.
A model-routing policy that picks the cheapest tier that can do the job and escalates only with a stated reason, expected value, budget, and fallback.
Polyglot delivery: Python and TypeScript at the core, with PowerShell, Go, JavaScript, and Shell across the portfolio.
A production PowerShell automation library with observability, security, and tests.
03 · Case Study: Systems over Goals
Stability is not a target you aim at. It is the residue of systems built before they are needed. ROFLcopter (a real-time multiplayer party word game, live in production) is the case in point: it stays up because keeping it up was engineered in, not chased after.
The product
A humor game that runs itself
Players race a set of random letters into the funniest phrase and vote on the winner. Empty seats fill with LLM bots, each writing in its own comedic voice; a panel of independent AI judges scores every round and learns what a crowd finds funny from real votes, storing no identities.
It never simply breaks. If the model key is missing, a call fails, or the batch has not warmed, the bots fall back to an offline part-of-speech generator. Graceful degradation is designed, not hoped for.
Deterministic by construction. A framework-free engine takes an injectable clock: production uses the real one; tests use a virtual clock that runs a full eight-round game plus the finale instantly, so even the finale is verified end to end with no device and no wall-clock wait.
Drop-proof rooms. A bot instantly takes the seat of anyone who leaves mid-game, so the room and the finale's jury never thin out because a player bailed. Abandonment is penalized, never rewarded.
Defensive parsing, taught by production. Models violate their own structured-output contracts. One wrapped its JSON in a markdown fence and silently dumped whole batches to the fallback, until the parser learned to unwrap fences, accept stray shapes, and line-split as a last resort.
The system that keeps it up
Capacity you measure, not capacity you guess
One box behind a single event loop is cheap to run and has an honest ceiling. The whole discipline is refusing to meet that ceiling for the first time during an outage.
The box reports on itself. Event-loop lag, the binding constraint of a single-process real-time server, is sampled continuously and charted against its ceiling on a token-gated owner panel, next to model-call latency, live rooms, and multi-week player and funnel trends.
A load harness, not a hope. A standalone headless harness ramps simulated rooms against production to measure the real capacity ceiling instead of guessing at it.
Config, not code. The moderation floor, the model identifiers, and the API keys are server configuration. Retuning the safety list or swapping a model is an operations task, not a release.
Boring deploys. A merge to main builds and ships on a self-hosted runner; a docs-only change deploys without restarting the live service.
The thesis, in one lineFrom the system's own white-paper: the capacity limit "is documented internally and drives the capacity posture rather than being discovered in an outage." Stability was never the goal. It is what the systems leave behind.
04 · The Judgment
The repositories show what I can build. This is how I decide what is worth building, and how I know when it is actually done. It is applied AI fluency: the four habits Anthropic names Delegation, Description, Discernment, and Diligence, learned the same way the technical skills were, by doing the work and paying for the mistakes.
Direction
Directing AI as a force multiplier
The defining skill of this stretch. I learned to run AI agents like a team: brief them clearly, let them work in parallel, and verify the output before I trust it.
Seen in a per-discipline generation swarm, a repeatable brief-building method, and the habit of walking finished work through real endpoints instead of reading a progress bar.
Taste
Knowing what to build, and what to leave alone
Product judgment is deciding which problems are real. I learned to tell correct behavior from a defect, and a business decision from a bug.
Seen in reading a "skipped" warm-up as correct drill placement, closing a list of audit defects that were mostly phantom, and treating a missing figure as a copyright call rather than a code fix.
Honesty
Calibrated confidence
I learned to separate what I have verified from what I assume, and to state which is which. The failure mode I most guard against is a confident claim built on an unchecked assumption.
Seen in a memory system that records my own corrections, catching myself over-claiming completion on database counts, and a standing rule to prove a claim by running it.
Economics
Spending like it is my own money
Every model call and every server hour is a real cost. I route work to the cheapest tier that can do the job and escalate only with a reason, an expected value, and a fallback.
Seen in a written model-routing policy and reuse-first design that cut a full program generation from $3.40 to under a dollar.
Ownership
Deciding, and carrying the call through
Working solo, I learned to make the decision, document it, and take it all the way to production without waiting for permission I do not need.
Seen in an operator doctrine that draws a clear line between what ships autonomously and the few irreversible calls that earn a second look.
Resilience
Steady under real operations
Production breaks. I learned to diagnose before touching anything, to pace load so systems stay up, and to recover without panic.
Seen in recovering a server from an out-of-memory hang, adding swap, and the rule to never stack a heavy generation, CI, and a deploy at the same time.
Communication
Translating across audiences and languages
The same work has to read one way to the engine, another to the buyer, and another to the person using it. I write the spec, the white-paper, and the product copy, and I build for English, Spanish, and Italian users.
Seen in a 40-page platform spec, a monetization architecture doc, published product briefs, and a personal writing voice I defined and keep.
Learning
Teaching myself, fast
The through-line under all of it. From a first app to production platforms in months, while building an engine whose whole job is how people learn.
Seen in the arc in section one, and a pedagogy layer that works out how a subject should be taught before it teaches it.
05 · The Portfolio Ledger
Every figure below is pulled live from the GitHub account. Commits and merged pull requests are exact.
This CV and white-paper's own source, versioned and backed up off-box
HTML
3
0
Live
regbak
Registry backup utility
Batchfile
2
0
Utility
Emotional_Event_Registry
Mood & event tracker, self-labeled "my first app"
HTML
11
0
Origin
06 · Operating Doctrine
Ten rules I did not have seven months ago. Most were paid for by a specific failure; all are now written into the operator doctrine I build with.
Shipping
Master is production
A directed merge to master is the live event, not a risk to flag. Take the task through the merge and deploy, then verify on prod.
Truth
Verify against reality
Running it beats reasoning about it. Walk the program through real endpoints, not database counts, and let behavior settle the question.
Grounding
Every generation is cited
Real corpus, real citations, zero fabricated identifiers. A model that sounds right is not the same as a model that is right.
Economics
Reuse before you regenerate
The cheapest call is the one you avoid. Catalog reuse cut a full program generation from $3.40 to under a dollar.
Method
Diagnose before fixing
Gather state, logs, and config first. No speculative commands. When a path is ruled out, it stays closed.
Vigilance
Hunt silent failures
Check the checks. A green pull request merged without branch protection can still take production down.
Completion
The unit of work is the system, not the diff
A task is done when the system is coherent and verified in the same session, not when the requested edit exists. Never optimize the green check at the expense of what it reports.
Migration
Every change is a migration
Rename, move, or replace something and you finish the transition in the same task. Two live paths with no marker is a defect, not a convenience.
Safety
Verify, then act
Every irreversible step (merge, deploy, delete, migrate) gates on a precondition that passes in the same run. Verify and act, or stop and report.
Control
Models are workers
The application owns policy, routing, credentials, memory, and final acceptance. Retrieved documents and tool output are untrusted evidence, never instructions.
07 · What Compounds Next
The platform is a factory now. Each new program costs less to build than the last, because the engine, the module system, and the money layer already exist. The work of these seven months turned prompting into architecture, and single apps into a platform that generates them.
Colophon. Every figure in this paper is drawn from the GestaltWorks GitHub account, read on 17 July 2026. Repository count, commit histories, and merged pull-request counts come from the GitHub API; project descriptions come from the repositories themselves; flagship velocity is measured on GestaltWorkframeEDU. Portfolio total: 1,761 commits and 580 merged pull requests across 16 owned repositories (one adapted fork excluded). Figures are conservative where an exact attribution to the seven-month window was not clean.
