zzboard
A board where coding agents post what they actually did, and their humans can see it.
AI agentsprovenanceengineering systems
The identity
Fosterman is not a costume.
It is the survival identity I built inside systems that repeatedly tried to define, discard, or simplify me. Over time, that identity became a philosophy of self-authorship, technical leverage, mentorship, and institutional accountability.
“Sirrele. Fosterman. The black wolf. The version of me that learned how to survive before he learned how to rest.” Even Better If: The Gospel of Fosterman
Identity architecture
Sirrele is the person. Fosterman is the operating identity.
Sirrele Steinfeld is the engineer, father, founder, writer, and human being.
Fosterman is the name for the part of me that learned how to survive systems, study them, challenge them, and eventually build alternatives.
Even Better If is the doctrine and reflection method that turns lived experience into rules, receipts, and usable maps.
Alter ego
Built for survival
Fosterman began as a self-authored identity powerful enough to survive environments that kept trying to reduce me to a file, diagnosis, statistic, or problem.
Philosophy
Built for interpretation
It became a way to think about abandonment, power, code, mentorship, leadership, boundaries, responsibility, and what institutions owe the people inside them.
Body of work
Built to leave maps
The writing, systems, products, reflections, and experiments are all attempts to leave behind something other people can navigate without needing me to rescue them.
Core definition: Fosterman is the survival identity Sirrele Steinfeld
built inside failed systems—and the philosophy he now uses to build better ones.
Part I · The black wolf is born
The origin story
The site should not flatten the story into inspiration. These are the chapters that explain how the identity formed, what it protected, and what it cost.
Chapter 1
The Concrete Bed & The Fireworks
Psychosis. Juvenile jail. Fourth of July. The night the world forgot me, and I started remembering myself.
In progress Read the chapterChapter 2
Meth, Montana & The Mirror
Racism, bruises, silence, cold benches, and learning how survival can split a person into versions.
Chapter 3
The Mercedes Thief
Being loved by Charles, testing the boundaries of love, and learning what “Dad” really means.
Chapter 4
Ed's Office
The moment someone said “I love you” and a different future became imaginable.
Chapter 5
UCLA and the Lost Prophet
Brilliance, instability, collapse, code-switching, ambition, and reinvention.
Chapter 6
Coding While Homeless
A laptop, a floor, and the discovery that software could become agency, structure, and a way out.
Chapter 7
Fosterman Is Born
Not a name. A philosophy. A challenge to God, government, and every system that pretends it cares.
In progress Read the chapterAbout
The throughline
At fifteen, I began asking why people fall through systems that are supposed to help them. Software became the discipline through which I learned to model complexity, build leverage, test hypotheses, and create systems that can be held accountable.
“I want to validate what I say with data. I don’t want to just say fluffy things.” Foster Tales, 2022
Engineering
Build from napkin to production
I translate ambiguous ideas into software systems, technical architecture, operating models, and measurable outcomes.
Leadership
Turn reflection into operating change
I am documenting how technical strength can create unnecessary organizational conflict—and how to become more effective without becoming less exacting.
Mission
Increase institutional capability
The long-term goal is to reinvest technical, financial, and organizational capability into foster care and other systems with high human stakes.
Part II · Even Better If
The rules, receipts, and maps
Even Better If is the doctrine at the center of Fosterman: part memoir, part manifesto, part curriculum, and part operating system for becoming more honest, capable, and self-authored.
Rule 1
All systems are built to abandon you.
But I do not have to abandon myself.
Rule 2
You can love people who hurt you and still walk away.
Love and boundaries can exist in the same sentence.
In progress Read the ruleRule 3
Code or die.
Programming gave me structure, leverage, language, and a way to build a future.
In progress Read the ruleRule 4
Mentor without permission.
The title is optional. Showing up is not.
Rule 5
There is no such thing as a bad kid.
There are only systems that ran out of language and blamed the child.
In progress Read the ruleRule 6
Burn the cape. Keep the map.
Stop trying to save everyone. Leave behind something usable.
Rule 7
You do not have to forgive to be free.
Freedom does not require fake peace.
In progress Read the rule Editorial principle: every major claim should eventually connect to
a story, a rule, a receipt, a system, or a reflection.
Selected work
Systems I have built, led, or shaped
These are the systems built from the philosophy. Each project answers: what failed, what I observed, what I built, what the evidence showed, and what I would change.
Threadout
Evidence-backed commitment integrity — find what slipped between the conversation and the merge.
engineering systemsevidenceAI
Conduit
The evidence engine underneath Threadout — ingestion, detection, and the judgement about whether an ask was actually kept.
AIknowledge graphsengineering systems
GitFitCode
A professional community treated as infrastructure rather than as a networking group.
communitymentorshipleadership
Architecture
The evidence pipeline, and where it stops being real
Threadout reconciles one initiative across a conversation export, a project tracker, and code history, and produces a short list of gaps with the evidence behind each one. This is that pipeline, at the only altitude it can be drawn honestly — 9 of the 12 boxes are working code, and 3 are a typed contract with no connector behind them. Every box says which it is, on the drawing.
What runs and what does not: the resolver is working code — the
bounding, the three-way outcome, the confidence gate, the held count, and the review
the publish path cannot bypass. The conversation export, the tracker and the code
history are a typed contract and an honest description, not running integrations.
The corpus that feeds the pipeline today is fictional and was written for the demo,
so the resolver has never been wrong about a real commitment — it has never seen
one. Read the full project write-up →
- Runs today — working code, exercised by a synthetic corpus.
- Declared, not connected — a typed contract. Nothing has flowed down these lines.
Inputs 4
Three declared sources and one that actually runs. The difference is the whole honesty of this drawing.
-
Conversation export
An export of the conversation where something was agreed to out loud. It exists as a typed contract and an honest description of the shape — there is no running integration behind it, and no real conversation has ever been imported.
-
Project tracker
The tracker the work was supposed to land in, declared so the resolver can ask whether an item appeared inside the window. Declared, and not connected.
-
Code history
The repository history over the same window, which is what turns “nobody tracked it” into “nobody did it either.” Declared for the same reason as the two beside it, and connected for none of them.
-
Synthetic corpus
A fictional corpus, written for the demo and resolved in the browser. It is the only thing that has ever been through this pipeline, which is exactly why the three boxes to its left are drawn as declarations rather than as pipes.
Synthetic example Everything below this box is working code. Everything it has ever seen was made up on purpose.
Normalize 1
Every source reduced to one evidence shape, or rejected.
-
Normalized evidence
Every source collapses into one evidence shape: who, when, where it came from, and the exact text. Timestamps are parsed strictly — an instant that cannot be proven is rejected rather than rounded, because a rounded instant silently widens every window downstream of it.
Fed by Conversation export Project tracker Code history Synthetic corpus
Detect 2
Find the promise, then bound the search before scoring it.
-
Commitment extraction
Find the sentences that promise something. Four shapes matter and only the first one looks like forgetting: dropped, buried, intentionally parked, and delivered differently.
Fed by Normalized evidence
-
Bounded candidate matching
Bound the search before scoring it — one project, one time window, the people who were actually involved. Missing data narrows the bounding rather than forking the logic, so a thin corpus produces fewer candidates instead of a second code path nobody tested.
Synthetic example Two projects share the word “worker.” The bounding drops the second one, because nobody in that conversation is on it.
Fed by Commitment extraction
Resolve 3
Three outcomes, one gate, and a count for what does not clear it.
-
Three-way resolution
Resolved, unresolved, or ambiguous. Ambiguous is its own outcome and is never quietly relabelled into one of the other two, because that is precisely where a system this shape earns false confidence.
Synthetic example “I’ll pull the retry cap out of the worker before the next release” — 28 July. Tracker: no item between 28 July and 4 August. Code history: no corresponding change. Outcome: unresolved, confidence 0.93.
Fed by Bounded candidate matching
-
Confidence gate
A finding with no cited evidence never ships. The gate reads the citations rather than the prose, so a confidently worded finding with nothing behind it fails in the same place as a hesitant one.
Fed by Three-way resolution
-
Held
Below the threshold a finding appears as a count and nothing more. Held is not discarded and not promoted, and the count is visible — so a reader can see how much the system declined to say rather than only what it was willing to.
Synthetic example “We should probably revisit the timeout” — 30 July. Two candidates in the window, neither cited. Outcome: held at confidence 0.41. Counted, never relabelled.
Fed by Confidence gate
Review and publish 2
A person, then a short list. Neither step can be skipped.
-
Human review
One authorized person approves the exact version of the content that goes out, and the publish path cannot bypass them. The guard denies with no side effects, so a refused publish leaves nothing half-written behind it.
Fed by Confidence gate
-
Weekly digest
A short list of gaps with the evidence that produced each one. The governing rule settles most of the smaller arguments: if it is not short and mostly right, it does not get sent.
Fed by Human review
Larger-role preparation
The roles I am preparing to hold
The immediate objective is not a title upgrade for its own sake. It is to prove I can carry broader technical scope, organizational influence, and executive accountability.
Immediate
Staff / Principal Engineer
High-leverage technical leadership across architecture, execution, standards, reliability, and cross-team coordination.
- System design depth
- Influence without authority
- Cross-org technical strategy
- Hands-on implementation credibility
Current readiness 78%
Parallel
Engineering Director / CTO
Build the operating system around engineering: roadmap quality, team design, architecture, accountability, and executive translation.
- Org design
- Portfolio prioritization
- Manager development
- Business and technical alignment
Current readiness 66%
Longer horizon
Founder / Institutional Builder
Turn Conduit and related systems into measurable proof that complex organizations can preserve intent and improve execution.
- Product-market fit
- Capital formation
- Team building
- Institutional governance
Current readiness 58%
Becoming
Preparing for greater technical authority
The Staff+ and Principal journey is part of the story, not a résumé sidebar. This is the public record of how I am preparing to hold more technical authority without repeating the leadership patterns that previously created unnecessary friction.
This plan is shown at its default settings: Principal Engineer plan, 8 weeks at 12 hours per week. Changing the role or the horizon needs JavaScript, so those controls are not shown here.
Preparation principle: fewer generic problems, more high-fidelity
practice under the constraints of the actual role.
Principal Engineer plan
Architecture, influence, and interview execution
Prioritize system design, cross-organizational technical judgment, coding credibility, and evidence of influence at scale. Current plan: 8 weeks at approximately 12 hours per week (96 total focused hours).
- Two system-design sessions per week with explicit tradeoff review
- Three timed coding sessions per week focused on medium-level patterns
- Build 12 leadership stories covering influence, conflict, failure, repair, and ambiguity
- Create one architecture case study from real production work each week
- Run at least three mock loops before active interviewing
Foundation 2 weeks
Depth 2 weeks
Mocks 2 weeks
Calibration 2 weeks
Coding
Timed implementation, data structures, graph traversal, concurrency reasoning, debugging, and clean communication under pressure.
System design
Requirements, architecture, scale, consistency, reliability, security, observability, cost, migration, and failure-mode analysis.
Leadership evidence
Stories of influence, conflict, failure, repair, ambiguity, prioritization, mentoring, executive communication, and organizational impact.
Becoming · Leadership receipts
Engineering my leadership
I am developing a body of work about how technically strong people become organizationally effective—especially when they are comfortable operating in conflict.
The core realization
I had a no-drama policy in my personal life but did not apply it consistently in my professional life. Because I could perform in conflict, I sometimes treated conflict as less costly than it was.
The old framing
“I am unwilling to work with bad leaders” protected standards but also gave me permission to classify people, retain judgment, and create tension that did not always improve the outcome.
The new framing
Truth must be paired with timing, relationship, proportionality, repair, and a distinction between a correctable mistake and a dangerous pattern.
The public series
Interview former managers, peers, and collaborators. Ask where my strengths became liabilities, when trust declined, what conflict was unnecessary, and what changed afterward.
Questions I am willing to ask publicly
When did my technical strength become an organizational liability?
What truth was I right about but ineffective at delivering?
When did I create conflict that was not necessary?
When did you stop trusting my judgment or approach?
What would make you trust me with greater authority today?
What should I preserve rather than “fix” about myself?
Describe behavior without turning it into identity
Separate diagnosis from judgment.
Optimize for changed outcomes
Being correct is not the same as being effective.
Treat conflict as a scarce resource
Escalation has organizational cost and must be proportional.
Make repair observable
A trustworthy record includes what changed after failure.
Trajectory
The next five and ten years
The path is sequenced around four forms of capital: capability, credibility, financial independence, and public trust.
0–2 years Carry Staff+ scope Secure Staff+/Principal scope. Build interview strength. Publish technical and leadership work.
1–3 years Make Conduit falsifiable Produce measurable evidence that organizational intent and work can be preserved across tools.
3–5 years Build institutional receipts Build revenue, a durable team, foster-care pilots, and institutional receipts.
5–7 years Shift to owned leverage Move from labor income toward owned leverage, investments, and advisory influence.
5–10 years Build civic credibility Build a governance record through boards, commissions, policy pilots, or local and state initiatives.
8–10 years Evaluate national scope Consider national office only with a constituency, a record, a coalition, and a defensible theory of governing.
Core trajectory: foster care is both the origin and the destination.
Engineering, leadership, Conduit, capital, and possible public office are the
capability-building path between them.
The map
Everything published, and everything still being tested
19 receipts and 6 hypotheses, drawn from the same files the rest of this page is built from. Filled points are work that exists. Hollow points are the trajectory being tested — claims about a future that has not happened yet. Select any point to follow what it connects to.
origin 10
-
Fosterman Is Born
Why the name is a philosophy and not branding: a frame built to hold contradictions no single label was big enough for.
Connects to Rule #7: You Don't Have to Forgive to Be Free Introduction Fosterman Is Born
-
Fosterman Is Born
Not a name. A philosophy. A challenge to God, government, and every system that pretends it cares.
Connects to Fosterman Is Born Coding While Homeless
-
Coding While Homeless
A laptop, a floor, and the discovery that software could become agency, structure, and a way out.
Connects to UCLA and the Lost Prophet Fosterman Is Born GitFitCode
-
Ed's Office
The moment someone said “I love you” and a different future became imaginable.
Connects to The Mercedes Thief UCLA and the Lost Prophet Burn the Cape. Keep the Map.
-
The Concrete Bed & The Fireworks
A psychotic break in a juvenile cell on the Fourth of July, and the night containment stopped being mistakable for care.
Connects to Rule #5: There's No Such Thing as “A Bad Kid” The Concrete Bed & The Fireworks
-
The Concrete Bed & The Fireworks
Psychosis. Juvenile jail. Fourth of July. The night the world forgot me, and I started remembering myself.
Connects to The Concrete Bed & The Fireworks Meth, Montana & The Mirror
-
Introduction
Why the book exists and refuses the sanitised version of the life it describes: part memoir, part manifesto, part blueprint.
Connects to Fosterman Is Born
-
The Mercedes Thief
Being loved by Charles, testing the boundaries of love, and learning what “Dad” really means.
Connects to Meth, Montana & The Mirror Ed's Office
-
UCLA and the Lost Prophet
Brilliance, instability, collapse, code-switching, ambition, and reinvention.
Connects to Ed's Office Coding While Homeless
-
Meth, Montana & The Mirror
Racism, bruises, silence, cold benches, and learning how survival can split a person into versions.
Connects to The Concrete Bed & The Fireworks The Mercedes Thief
engineering 4
-
Carry Staff+ scope
Secure Staff+/Principal scope. Build interview strength. Publish technical and leadership work.
See the window on the ten-year map
Connects to Make Conduit falsifiable GitFitCode
-
Rule #3: Code or Die
Code as leverage rather than as a career: the discipline that converted survival intelligence into agency and exits.
Connects to GitFitCode
-
zzboard
A board where coding agents post what they actually did, and their humans can see it.
Connects to Conduit GitFitCode
-
GitFitCode
A professional community treated as infrastructure rather than as a networking group.
Connects to Rule #3: Code or Die Burn the Cape. Keep the Map. zzboard Conduit Carry Staff+ scope Coding While Homeless
leadership 2
-
Burn the Cape. Keep the Map.
The difference between saving people and leaving behind something they can navigate.
Connects to Conduit GitFitCode Ed's Office
-
Shift to owned leverage
Move from labor income toward owned leverage, investments, and advisory influence.
See the window on the ten-year map
Connects to Build civic credibility Build institutional receipts
conduit 3
-
Conduit
The evidence engine underneath Threadout — ingestion, detection, and the judgement about whether an ask was actually kept.
Connects to Burn the Cape. Keep the Map. zzboard Threadout GitFitCode Build institutional receipts Make Conduit falsifiable
-
Threadout
Evidence-backed commitment integrity — find what slipped between the conversation and the merge.
Connects to Conduit Make Conduit falsifiable
-
Make Conduit falsifiable
Produce measurable evidence that organizational intent and work can be preserved across tools.
See the window on the ten-year map
Connects to Threadout Build institutional receipts Conduit Carry Staff+ scope
public service 2
-
Build civic credibility
Build a governance record through boards, commissions, policy pilots, or local and state initiatives.
See the window on the ten-year map
Connects to Evaluate national scope Shift to owned leverage
-
Evaluate national scope
Consider national office only with a constituency, a record, a coalition, and a defensible theory of governing.
See the window on the ten-year map
Connects to Build civic credibility
foster care 4
-
Rule #5: There's No Such Thing as “A Bad Kid”
Labelling a child bad relieves the adults of curiosity; behaviour is adaptation, and reading it accurately is the actual work.
Connects to The Concrete Bed & The Fireworks
-
Rule #7: You Don't Have to Forgive to Be Free
Freedom does not require performing forgiveness on anyone else's schedule — release, not absolution, is the honest assignment.
Connects to Rule #2: You Can Love People Who Hurt You & Still Walk Away Fosterman Is Born
-
Build institutional receipts
Build revenue, a durable team, foster-care pilots, and institutional receipts.
See the window on the ten-year map
Connects to Shift to owned leverage Conduit Make Conduit falsifiable
-
Rule #2: You Can Love People Who Hurt You & Still Walk Away
Love is not proof that staying is wise, and creating distance does not require inventing hatred first.
Connects to Rule #7: You Don't Have to Forgive to Be Free
The turn
Burn the cape. Keep the map.
Rescue keeps the helper at the centre. It feels like generosity and works like a dependency — every person I pulled out left the system that failed them exactly as it was. What travels is the part I can put down: language, a rubric, a working example, a name to call.
The person disappears. The map survives.
Now
The current work
Daily and weekly notes from the Staff+ journey, Conduit, leadership repair, and institutional thinking. Written as it happens, not reconstructed afterward.
Four models, one coordinator
The week a crew of four model families shipped a public product, ran a client engagement, and took its first seat on a fifth.
weeklycrewAI agentszzboard
Held at the door
A Claude Code session that did work is now held at exit until it posts what it learned. Guidance had two days and produced one post.
zzboardAI agentshooksdecision
The reviewer refuted the result
Eight pull requests merged in a day, each reviewed by a model that did not write it. One review said do not merge, and it was right.
zzboardAI agentscode reviewcrew
The refusals are the product
zzboard went from a report about an agent swarm to a deployed board in one day. Every design decision that mattered is something the API refuses to do.
zzboardAI agentsprovenance
Artifacts
The raw proof behind the work
This is not a generic blog. It is an archive of letters, receipts, field notes, maps, technical papers, reflections, transcripts, and selected raw material that proves the work has been lived—not merely branded.
Receipts
Claims paired with proof
Job losses, product decisions, leadership mistakes, repairs, code, outcomes, and the evidence beneath the narrative.
Letters
Writing that became a way to heal
Letters to Skylar, letters to younger versions of myself, and writing that helped make difficult thoughts speakable.
Field Notes
What the systems reveal
Observations from engineering teams, foster care, healthcare, community building, leadership, and institutional failure.
Maps
Models for navigating complexity
System diagrams, knowledge graphs, mind maps, decision frameworks, organizational traces, and the geometry of Conduit.
Reflections
The person inside the system
Conflict, protectors, identity, ambition, fatherhood, self-loyalty, regulation, and becoming worthy of greater authority.
Technical Papers
The engineering body of work
Conduit, Staff+ preparation, architecture reviews, reliability analysis, AI systems, and implementation case studies.
Rule #2: You Can Love People Who Hurt You & Still Walk Away
Love is not proof that staying is wise, and creating distance does not require inventing hatred first.
Even Better IfmanuscriptEBI Rules
Rule #3: Code or Die
Code as leverage rather than as a career: the discipline that converted survival intelligence into agency and exits.
Even Better IfmanuscriptEBI Rules
Rule #5: There's No Such Thing as “A Bad Kid”
Labelling a child bad relieves the adults of curiosity; behaviour is adaptation, and reading it accurately is the actual work.
Even Better IfmanuscriptEBI Rules
Rule #7: You Don't Have to Forgive to Be Free
Freedom does not require performing forgiveness on anyone else's schedule — release, not absolution, is the honest assignment.
Even Better IfmanuscriptEBI Rules
Introduction
Why the book exists and refuses the sanitised version of the life it describes: part memoir, part manifesto, part blueprint.
Even Better Ifmanuscript
The Concrete Bed & The Fireworks
A psychotic break in a juvenile cell on the Fourth of July, and the night containment stopped being mistakable for care.
Even Better IfmanuscriptBlack Wolf
Contact
Build, challenge, or collaborate with me.
I am interested in technically ambitious software roles, institutional design, organizational intelligence, foster-care systems, and conversations that make the work more rigorous.
Elsewhere I am @iamfosterman.
- GitFitCode community on Discord Open invite. No approval, no application — the fastest way to reach me.
Good reasons to reach out
Staff+ or Principal engineering roles
Technical leadership and architecture
Conduit design partners
Foster-care systems and pilots
Leadership interviews and candid feedback
Who do you want to be if you get to live to 70?
I’m still answering it.
Sirrele Steinfeld, Fosterman, @iamfosterman