Mark G.

Lead SWE

Software development

Solid coverage

Before your AI advises you on software development, it reads how Mark has actually called it — the decisions, the reasons, and the lines they will not cross.

Use this model for

  • Backend Data Design
  • API and Dependency Selection
  • CI/CD and Release Scope
  • Testing Workflows
  • AI Application Architecture
  • Access Control

When Mark’s model can help

Mark has not published a worked example yet. What is below is what their model actually holds — the decisions themselves, in their words.

Copy an example prompt4 patterns with a gap to fill, once you are connected

  • Before you commit

    A full pass over a plan or a draft, with the changes it needs.

    Use Mark G.'s Decision Model on avva before we lock this in. What we're deciding: <paste your plan, doc, or diff here> Run their review, follow the returned protocol exactly, and give me the verdict and any required changes before you continue.

  • Choosing between options

    A call between two options, their hard stops checked first.

    Use Mark G.'s Decision Model on avva to compare these. Option A: <describe it> Option B: <describe it> Context: <what constrains the choice — API and dependency selection> Check their hard stops first, then compare the options under their judgment, and tell me which one they would have taken and why.

  • A quick gate

    A cheap scan against the lines they will not cross.

    Run this past the hard lines in Mark G.'s Decision Model on avva before I send it. <paste the draft, plan, or change> Tell me only what they would refuse and why. If nothing trips, say so.

  • Before you have an approach

    The questions they ask before an approach gets locked in.

    I'm starting on backend data design. Ask Mark G.'s Decision Model on avva what they would want settled first. Situation: <two or three lines> Give me their questions, and flag the ones my situation does not answer yet.

Patterns with a gap to fill, not examples of what Mark answered. The one worked answer on this page is the one they approved themselves; anything else here would have been written by us, under their name.

Connect Mark G. to your AI

Three steps. The first works right now with no key or account; the last is the one that makes your AI consult Mark G. without being asked.

  1. https://avva.chat/mcp/software-development-mark-g

    Customize → Connectors → + → Add custom connector, or the + in the chat box → Connectors → Manage connectors. Missing on a company-managed account? An admin holds it — Claude Code, Cursor, and Gemini CLI take a local config instead.

    This returns a smaller slice — the rules and cases this expert puts first, 40 consults a day, per model — enough to see how this expert reasons and whether their judgment fits your problem. Step 2 returns the rest.

    Check it worked. Send this in the chat once the connector is in.

    Ask Mark G.'s Decision Model on avva what kinds of decision it covers, and summarize it for me in three lines.

Not legal, medical, financial, or other professional advice. Terms of Use

About this model

Built from 48 approved decisions across AI chat history, code & reviews and issues & roadmaps. Updated September 13, 2026. Every rule below is counted against decisions Mark approved — open any of these to check the grounding yourself.

How this model is groundedWhat it holds, and what Mark weighs first
decisions
48
forces
16
criteria
16
stop rules
19

What Mark weighs first

Stable backend behavior across asynchronous updates and schema changesMUST WEIGH
An explainable reaction lifecycleMUST WEIGH
Reproducible source-to-production deliveryMUST WEIGH

3 of 16 criteria — the ones the model puts first. The other 13, what each one looks for, the stop rules and the switches arrive when your AI asks. Without an account that is a smaller slice — the rules and cases this expert puts first. Signing in returns the complete published model.

Under 5 named conditions Mark’s usual answer flips. Those switches, and 19 lines they will not cross, go to your AI with the rest.

See examples of approved decisions2 of 48, in Mark’s own words

The first 2 in the record, in published order — not the ones nearest to your situation, and not a summary of the rest. Your AI does not repeat a call it reads here: it checks what settled that case against your own facts, and weighs it against which of Mark’s forces beat which across all 48. Signed in, your AI receives all 48, ordered against yours.

  1. The situationA professional decision-model platform lets experts collect material in a desktop app and makes their models available through MCP and a marketplace. The collection algorithm must choose among the source's available tools.

    The callUse a safe read-only tool first. If none is available, use the tool provided, but use it only to read.

  2. The situationA profile-feed application's backend handles user reactions; a separate index updater consumes NATS messages and writes to SQL and Elasticsearch. A minimal test-environment file is being introduced for CI unit tests, along with proposed changes to application code.

    The callKeep tests_env as a file for CI. Do not change application code for it; unlike env_example, it should contain only what unit tests need.

    WhyThis is for CI.

Technical methodologyHow 24 recorded contests become one order
  1. 48real decisions, in their own words — from AI chat history · code & reviews · issues & roadmaps
  2. 16forces those decisions turned on — the pressures Mark decides under
  3. 24head-to-head contests: every case records which force won and which lost
  4. One orderwhich force wins when two collide — read off those contests, never ranked by hand

Nobody types a ranking in. How it works walks through where the order comes from.

How Mark decides, in summaryOne paragraph, written from the approved set

Consult this model for software architecture, backend data changes, development workflows, and release scope. Its cases preserve concrete calls on compatibility, inspectable state, reproducible delivery, expert-controlled AI workflows, and deliberately limited prototypes.

Their experience is above. This describes what this particular model encodes — the dispositions its rules were built from, not a claim about everything Mark knows.

What’s trustworthy here
  • Identity: Named on the public page · Lead SWE. Self-reported, not verified. Judge by the record on this page and by the packet your AI follows.
  • Scope: Software development — Backend Data Design, API and Dependency Selection, CI/CD and Release Scope, Testing Workflows, AI Application Architecture, Access Control, Developer Tools
  • Coverage: A solid set of decisions behind its rules, with repeated patterns — usable day to day.
  • Privacy: Their cases and rules, in their own words. Client names, employer names, and source quotes are stripped before storage.
  • Whose answer it is: avva runs no AI of its own. Your AI reads this model and answers you, and it names where the model had nothing to say.
  • Limits: Informational tooling for agents, not professional advice. avva is provided as-is.

Privacy Center · Terms

Know someone whose calls you would want your AI to have? Send them to Studio — publishing is free during beta.