# How to build a personal evidence garden

## What it is for

A personal evidence garden is a private, growing record of work you have actually done and the evidence that supports it.

It is especially useful when your experience does not form a tidy career ladder. You may have worked across several fields, started ventures, contributed to other people’s projects, pursued serious interests outside paid employment, or repeatedly become the person who solves a certain kind of problem without ever receiving the corresponding job title. A conventional CV compresses all of this. Sometimes it makes a highly capable person look strangely blank.

An evidence garden reverses the process. Instead of starting with a polished claim such as “I am a strategic thinker”, it starts with concrete episodes:

- a difficult problem you helped solve;
- something you built, repaired, organised or improved;
- a decision for which you carried responsibility;
- a situation in which other people relied on your judgement;
- an idea you turned into a practical method, product or activity;
- work that mattered even though it was informal, voluntary, interrupted or never made public.

Over time, patterns may become visible across those episodes. The garden can then help with career exploration, CVs, applications, interviews, proposals, biographies or portfolios. But those are later uses. Its first purpose is to recover the underlying evidence.

## It is not a peacocking exercise

Recovering evidence and promoting yourself are different tasks.

You do not need to decide that an episode is impressive before recording it. You only need to describe what happened accurately enough that it can be examined. The record can include uncertainty, mixed results, shared credit, missing documents and things you would now do differently.

This is particularly helpful for people who dislike self-promotion. The question is not, “How can I make myself sound exceptional?” It is, “What did I notice, decide, make, change or contribute, and what traces of that work still exist?”

Accurate description may reveal substantial ability. That is allowed. Modesty does not require making the evidence disappear.

## Why ordinary career records are often insufficient

CVs, LinkedIn profiles and short biographies are presentation formats. They show selected conclusions but usually omit the material from which those conclusions were drawn.

They are particularly poor at preserving:

- work that crossed the official boundaries of a role;
- judgement, diagnosis and problem framing;
- contributions made inside a team;
- experiments and unsuccessful ventures that still demonstrate capability;
- connections between work done in different fields;
- responsibilities that were real but never formally named;
- the scale, constraints and complexity surrounding a piece of work;
- supporting documents and the precise source of a claim.

People with ADHD or wide-ranging interests may face an additional problem: the work exists, but it is distributed across old files, changing interests, forgotten conversations and other people’s projects. Memory is expected to function as an archive. It is not especially good at the job.

## What counts as evidence

Start broadly. Possible sources include:

- CVs and old applications;
- proposals, reports and presentations;
- products, websites, tools, photographs or other things made;
- emails, messages, meeting notes and feedback;
- testimonials and recommendations;
- publications, talks and teaching materials;
- records of voluntary, community or self-directed work;
- your own recollection;
- another person’s recollection of working with you.

These sources do not all have the same status. A dated report can establish that an output existed. Your memory may explain the decisions behind it. A testimonial shows what someone else observed, but it does not prove every surrounding claim. The garden should preserve these distinctions rather than flattening everything into “proof”.

The same story repeated across several CVs is useful for recovering context, but it is not several independent confirmations.

## The basic unit: a bounded episode

The most useful unit is usually a particular episode rather than a broad skill.

“Good at strategy” is a claim. “Took a confused client brief, identified the decision it needed to support and designed a research approach that made the question answerable” is an episode that can be investigated.

An episode record might include:

- **Stable ID:** a short identifier that will not change when the title or categories change.
- **Plain title:** enough to recognise the episode without turning it into advertising copy.
- **Summary:** what happened in a few sentences.
- **Context:** when, where and under what conditions the work happened.
- **The problem:** the difficulty, need, constraint or opportunity.
- **Personal contribution:** what the person actually noticed, decided, made, changed or carried.
- **How the work connected:** how that contribution related to the wider process or outcome.
- **Outputs and traces:** documents, products, decisions, feedback, scale or observable results.
- **Possible capability links:** cautious interpretations suggested by the episode.
- **Sources:** where the supporting information came from.
- **Gaps and confidence:** what remains uncertain or depends on memory.
- **Confidentiality:** what may be stored privately but should not be published.

Capability links should remain provisional. A label derived from one story is a lead. A pattern seen across several different episodes is more informative.

## A practical process

### 1. Map the territory without forcing a life story

Make a loose inventory of roles, periods, projects, ventures, collaborations, interests, communities and less official work. Include things that were important even if they never became a job title or source of income.

Do not try to make the history coherent yet. A garden is allowed to start as a collection of oddly shaped plots.

### 2. Choose one promising thread

Start where either memory or surviving material is reasonably rich. This could be one role, one venture, one type of problem or a project that other people often mention.

Three well-recovered episodes are more useful than an exhausting attempt to inventory an entire career at once.

### 3. Use a conversation rather than a large form

An interview works better than asking someone to complete a comprehensive questionnaire. Broad prompts such as “What are your achievements?” create both cognitive load and an immediate demand to perform importance.

Ask one or two concrete questions at a time. Follow the details that appear. Useful prompts include:

- What was happening when you became involved?
- What was unclear, broken, difficult or missing?
- What did you notice that others had not yet noticed?
- Which decisions were actually yours?
- What did you make or change?
- What constraints were you working around?
- Who else was involved, and how did your contribution differ from theirs?
- What existed afterwards that had not existed before?
- Is there a document, object, message or person that could help reconstruct this?
- Which part of this account are you least certain about?

### 4. Inspect the surviving traces

Add relevant documents and compare them with the remembered account. Documents often recover dates, scale, wording, collaborators and outputs. Conversation often recovers the invisible work: judgement, uncertainty, negotiation and why a decision made sense at the time.

Neither source should automatically dominate the other. The point is to build a more inspectable account.

### 5. Draft cautiously

Separate personal contribution from team and organisational outcomes. Do not attribute a whole project to one person merely because their name appears in the material. Do not invent a result because a proposal described an intention.

Mark interpretation as interpretation. Keep open questions visible. A useful draft is not the same as a verdict.

### 6. Assign stable identifiers

Give each episode and source an identifier, such as `EP-001` and `S01`. This allows people and AI tools to refer to an exact record even if its title, wording or categories later change.

Identifiers also make correction easier: “EP-004 overstates my role” is less ambiguous than “that story about the workshop”.

### 7. Look for patterns only after several records exist

Once there are three to five records, examine them side by side. What kinds of problems recur? What forms of contribution appear in different settings? What do people repeatedly entrust to this person? Which conditions seem to bring out their best work?

Use the emerging patterns to propose ways of browsing the garden. These might concern types of problem, modes of contribution, environments, outputs, domains or practical questions the evidence can answer.

Do not import somebody else’s categories merely because they worked for that person.

### 8. Build the smallest useful system

A first version might contain:

- five to ten reviewed episode records;
- a source register;
- a place for incomplete leads;
- search or simple topic filters;
- a few questions that retrieve relevant episodes.

A private website is convenient because it makes the material scannable and easy to revisit. It is not essential. The same method can start in Markdown files, a table or a notes system. The evidence model matters more than the furniture.

## Starting brief to give ChatGPT

Copy the text below into a new ChatGPT conversation. Attach any useful starting documents if you have them, but it is also possible to start from memory.

> Help me build a private evidence garden: a structured, revisable record of work I have actually done and the evidence that supports it. This is not a request to write a polished biography, personal brand or public portfolio.
>
> My career may be nonlinear and spread across different roles, interests, paid work, voluntary work, projects or periods of self-directed learning. Do not force it into a conventional career ladder or start with a fixed list of skills. Help me discover the recurring capabilities and kinds of problems through concrete episodes.
>
> Interview me gradually. Start by mapping the broad territory: roles, projects, ventures, collaborations, difficult problems, things I built or improved, responsibilities people trusted me with, and work that others may not have seen. Then choose one promising episode at a time. Ask specific questions about:
>
> - the context and what was happening;
> - the problem, constraint or opportunity;
> - what I personally noticed, decided, made, changed or contributed;
> - how my contribution related to the wider team or outcome;
> - outputs, scale, feedback or other observable traces;
> - documents or people that could support the account;
> - what is uncertain, confidential or based only on memory.
>
> Do not make me perform confidence or sell myself while we are recovering the evidence. Do not mistake modest wording for lack of ability. If I give a vague statement, ask for an example rather than turning it into a grand claim. Separate facts, my recollection, other people’s statements and your interpretation. Never invent outcomes or attribute a team result wholly to me.
>
> For each usable episode, draft a record with a stable ID, a plain title, short summary, context, problem, personal contribution, how the work connected, outputs or traces, possible capability links, source references, confidence or gaps, and confidentiality notes. Treat capability links as hypotheses that can change as more episodes appear.
>
> After three to five records, show me the patterns you can cautiously see and propose useful ways to browse them, such as recurring problems, types of contribution, settings or questions the evidence can answer. Let the structure emerge from my material. Keep the collection private unless I explicitly decide what to publish.
>
> Start by explaining the process briefly, then ask me one manageable opening question. Do not give me a large questionnaire all at once.

## What happens after the interview starts

The first conversation does not need to produce a finished career architecture. Its job is to recover a small amount of reliable material and establish a way of working.

Review each drafted episode. Correct overstatement, missing context and falsely tidy causality. Add sources gradually. Keep incomplete leads rather than forcing them into finished records. Once the collection contains enough varied examples, decide whether a private searchable site would make it more useful.

Only after that should you consider what, if anything, belongs in public. The private garden can remain much richer, messier and more honest than the versions assembled for particular audiences.