---
title: "Stop Re-Explaining Your Business to AI"
description: "Start your agent from your business record instead of another reconstruction session, with one inspectable context path and clear approval rules."
canonical: "https://scalewithsearch.com/articles/stop-re-explaining-your-business-to-ai"
date: "2026-08-17"
modified: "2026-09-25"
---
## Site navigation

- [Scale With Search](https://scalewithsearch.com/)
- Services
  - Local
    - [All services](https://scalewithsearch.com/services)
    - [AI visibility tracking](https://scalewithsearch.com/services/ai-visibility-tracking)
    - [Google Maps rank tracking](https://scalewithsearch.com/services/google-maps-rank-tracking)
    - [Local SEO](https://scalewithsearch.com/services/local-seo)
    - [Google Business Profile management](https://scalewithsearch.com/services/google-business-profile)
  - Website and search
    - [SEO solutions](https://scalewithsearch.com/seo)
    - [Technical SEO](https://scalewithsearch.com/seo/technical-seo)
    - [On-page and content SEO](https://scalewithsearch.com/seo/on-page-and-content-seo)
    - [Semantic SEO and topical authority](https://scalewithsearch.com/seo/semantic-seo-and-topical-authority)
    - [Custom SEO tools](https://scalewithsearch.com/seo/custom-seo-tools)
    - [Site build](https://scalewithsearch.com/work#site-build)
    - [Whole website fix](https://scalewithsearch.com/work#whole-website-fix)
  - AI functions
    - [AI operations](https://scalewithsearch.com/ai-systems)
  - First step
    - [Working session](https://scalewithsearch.com/audit)
  - Start here
    - [Send your brief](https://scalewithsearch.com/work#send-your-brief)
    - [Example scope](https://scalewithsearch.com/work#review-the-example-scope)
    - [Prices and terms](https://scalewithsearch.com/work#prices-and-terms)
- Industries
  - Agencies
    - [SEO agencies and consultants](https://scalewithsearch.com/for/seo-agencies-and-consultants)
    - [Marketing agencies and freelancers](https://scalewithsearch.com/for/marketing-agencies-and-freelancers)
    - [AI visibility check for SEO agencies](https://scalewithsearch.com/check/seo-agencies-and-consultants)
    - [Marketing and SEO agencies](https://scalewithsearch.com/for/#group-agencies)
  - Browse by business
    - [Trades and field service](https://scalewithsearch.com/for/#group-trades)
    - [Real estate agents and teams](https://scalewithsearch.com/for/real-estate)
    - [Health, wellness, and animal care](https://scalewithsearch.com/for/#group-practices)
    - [Chiropractic and wellness practices](https://scalewithsearch.com/for/chiropractic-and-wellness-practices)
    - [Law, finance, and IT](https://scalewithsearch.com/for/#group-professional)
    - [Machine shops and manufacturers](https://scalewithsearch.com/for/#group-manufacturing)
    - [Restaurants, shops, studios, and nonprofits](https://scalewithsearch.com/for/#group-local-services)
  - [All industries](https://scalewithsearch.com/for/)
- Resources
  - Free tools
    - [AI visibility check](https://scalewithsearch.com/check)
    - [All free tools](https://scalewithsearch.com/tools/)
    - [Free starter kit (.zip)](https://scalewithsearch.com/kit/business-memory-starter-kit.zip)
  - Guides
    - [AI-native library](https://scalewithsearch.com/ai-native)
    - [Industry intelligence](https://scalewithsearch.com/intelligence)
    - [Office job guides](https://scalewithsearch.com/guides/)
    - [Business memory](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide)
    - [Agent governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business)
    - [Ownership and migration](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide)
  - Articles
    - [All articles](https://scalewithsearch.com/articles/)
- About
  - Evidence
    - [Proof](https://scalewithsearch.com/proof)
  - Scale With Search
    - [About](https://scalewithsearch.com/about)
    - [How it works](https://scalewithsearch.com/how-it-works)

# Stop Re-Explaining Your Business to AI.

Synthetic example, not a client result: you open an agent to prepare Monday's client update. Before it can help, you paste the current scope, the client's preferred format, the last decision, and the work that must stay out of the update. You did the same thing last Monday.

The problem is not that the model forgot one detail. The work has no durable briefing path.

A useful agent needs a source record before it needs a clever instruction. That record can be small. It must be controlled by you, readable by the system, and specific enough to constrain the task.

The [homepage's synthetic example](/#worked-example) shows the same failure at sentence level: a customer update drafted without the files claims a wider service area, promises a finish date nobody set, and plans a call the customer did not ask for. Checking the source files restores the service area, the open finish date, and the texts-only rule. This article builds the record that makes that check repeatable.

## Reconstruction is hidden work

Repeated explanation feels like normal AI use because it happens inside the chat box. It is still labor.

You search old messages. You decide which facts remain current. You restate exceptions. You warn the agent about a mistake it made last time. Then you ask for the output you wanted in the first place.

That is two jobs:

1. reconstruct the operating context;
2. produce the requested work.

The first job should not depend on what you remember at the moment. It should read from a maintained record.

Chat exports can preserve conversations, but an archive is not yet a working briefing system. OpenAI's official export process returns a ZIP that includes chat history and other account data. That is useful for custody and migration. It does not decide which client facts, rules, and corrections apply to today's task. [OpenAI: export ChatGPT history and data](https://help.openai.com/en/articles/7260999-how-do-i-export-my-chatgpt-history-and-data)

## Count what the repetition costs

The repetition is not a prompting failure. Language models keep no state between sessions unless a product feature or a file puts it back. Product memory features keep fragments, such as your name, your industry, or a pet's name, not the operating context that makes output fit your situation.

The visible cost is time: minutes of setup in every session before the real question, repeated each working day.

A worked example assumes ten minutes of setup per conversation and three conversations a day. That is 30 minutes a day, 2.5 hours a week, and 125 hours across a 50-week year. Over five years it is 625 hours. Multiply the hours by your own hourly rate.

The ten minutes is an assumption for the example, not a measured average. Time your next three sessions and use your own figure. At 5 to 15 minutes per session and 10 to 20 sessions a week, the range runs from under one hour to five hours a week.

For a proposal, the ten minutes often splits into three parts. Minutes one to three cover the services, the pricing model, and the client type. Minutes four to six cover brand voice and tone. Minutes seven to ten cover the specific client and scope. Only then does the proposal start.

The hidden cost is quality. When setup takes effort, people cut corners. They summarize the context instead of stating it, and they skip details that feel tedious to type again. The model fills each gap with a generic assumption, so the output turns generic too.

The third cost is inconsistency. A different explanation in each session produces a different interpretation. Monday's draft and Friday's draft disagree because the context differed, not because the question did.

Three more costs follow. The setup is tiring, so it drains attention before the real work begins. People stop using AI for quick tasks, because ten minutes of setup outweighs a two-minute job. And the owner becomes a bottleneck. Staff cannot delegate AI work when only the owner knows how to explain the context.

Other business tools treat memory as baseline. The CRM keeps contacts. The calendar keeps the schedule. The mailbox keeps the correspondence. An AI workflow without a maintained record asks a person to supply that baseline by hand.

## See why the usual workarounds stall

Most teams try four workarounds before they build a record:

- **Opening paste.** A block of context goes at the top of each chat. It works until someone forgets one detail and the output goes wrong.
- **Custom instructions.** When ChatGPT launched them in 2023, each field held 1,500 characters, about 250 words. A business needs more than that.
- **Projects and custom GPTs.** They hold more material, but the context stays inside one product and does not follow a switch to another tool. The [Claude Projects and custom GPTs comparison](/articles/claude-projects-vs-custom-gpts-business-memory) tests both.
- **Saved prompts in notes.** The prompt still needs a manual paste in every session.

Each workaround accepts the reset and asks a person to work around it forever. A working record has three properties instead. It is a structured knowledge base of business facts, processes, clients, voice, and terms. The agent reads it at the start of each session without a request. It persists, so a new client added once appears in every later session. The [Obsidian and Claude connection guide](/articles/connect-claude-obsidian-business-memory) shows one way to hold that record.

A first version takes a few hours. After that, output quality stops depending on how thoroughly someone explained the business that day.

## Start with one recurring incident

Do not begin by documenting the entire company. Pick the recurring job that causes the most reconstruction.

Good starting jobs have a visible finish line:

- prepare a client status brief;
- draft a proposal from approved scope;
- summarize a weekly operating review;
- classify incoming requests;
- prepare research for a decision.

Write down what you explain every time. Separate source facts from task instructions.

The source facts may include the client, price, approved scope, exclusions, current status, and prior decisions. The task instructions may include the requested output, due date, source order, review method, and stopping point.

That separation matters. A price change belongs in the business record. "Prepare Friday's update" belongs in the task brief.

## Write down what you repeat

The repeated material is usually specific. One owner's list covers name, role, clients, tone, proposal structure, and pricing tiers. It also names three clients under NDA who must never appear in examples. A house rule adds no Oxford comma in headlines but an Oxford comma in body copy.

A video editing service for SaaS companies shows the pattern. On Monday the owner types: "I run a video editing service for SaaS companies. We charge a flat monthly fee for four videos. Our turnaround is five business days. Write a proposal for a client who needs product demo videos." On Tuesday, the same three sentences open an email to a prospect who went quiet after a discovery call. A week later, they open a LinkedIn post about a new client review. The context never changed. Only the task did.

Move those sentences into the record once. A context file for this business covers four areas:

- company: services, prices, packages, and constraints;
- clients: industries, personas, and NDA restrictions;
- voice: tone, banned phrases, and sentence structure;
- tools: the CRM, platforms, integrations, and workflows.

Write the entries as rules, not prose:

```text
Voice: No fluff. No "Let me unpack that." Use contractions. Vary sentence length.
       Short sentences for impact. No wrap-up conclusions.
Pricing: flat monthly package for 4 videos, 5-day turnaround; rush fee for 48-hour delivery;
         no equity deals, no trade, no payment plans below the minimum project value.
Clients: B2B SaaS only. No ecommerce, personal brands, or local businesses.
         Current clients are under NDA. Never name them in examples.
Email: no pushy language; include the calendar link in every outreach.
```

With that record in place, "Write a proposal for the Acme Corp opportunity" needs no preamble. In Claude Code, a `CLAUDE.md` file in the working directory loads at session start. The agent can also read client briefs, past proposals, templates, and meeting notes from the same folder, such as an Obsidian vault. The fix is a file. The work is to document the business clearly enough that a model can apply it.

## Compare one request with and without the record

The repetition is not limited to new days. It returns when you move from phone to laptop and when you open a different AI tool.

Without a record, a simple request becomes an interview:

> You: "I need help drafting an email to the Acme Corp client about the project delay."
>
> AI: "Can you tell me more about the project and your relationship with this client?"
>
> You: "Acme Corp is our enterprise client. We are three weeks into a six-month engagement. The delay is because their team has not provided the data we requested. I need to escalate politely but firmly."
>
> AI: "What is your typical tone with this client?"
>
> You: "Professional but friendly. I have worked with their VP of Operations for two years. He appreciates directness."

Six messages of background come before the first draft. With the record, one message is enough: "Draft an email to Acme Corp about the project delay due to missing data." The agent reads that Acme Corp is an enterprise client, reads the relationship history and tone preference, and drafts.

A first file for this purpose can hold six kinds of entry, in bullet points or partial sentences:

- identity: your name, role, company, what you do, and who you work with;
- current projects: each by name, with a one-line description;
- processes: the high-level steps of common workflows, such as "Client onboarding: contract review, kickoff meeting, access setup, first deliverable";
- preferences: email format, summaries or raw data, and client tone versus internal tone;
- tools: the names of the CRM, email system, and project management software;
- domain knowledge: industry terms, company acronyms, product names, and team members with their roles, or anything you would explain to a new employee.

Keep it in one folder, such as `AI Context`, as `CLAUDE.md`, `context.md`, or `memory.md`. The name is a convention, except that Claude Code looks for `CLAUDE.md`. When one file grows, split it into `clients.md`, `processes.md`, and `preferences.md`, and point to them from the main file. When a project starts or a process changes, the update takes about two minutes.

The file is not a long system prompt. Claude Code loads `CLAUDE.md` at session start, and it opens the other files when the task needs them. ChatGPT does not read local files, so there the record goes in through custom instructions, within their length limit, or a paste at the start of a complex conversation.

The larger gain is attention. You stop tracking what you told the AI and in which session. The [time measurement guide](/articles/how-much-time-wasted-on-ai-context) shows how to log the minutes before and after the file exists.

## Give the agent a path, not a pile

A folder can be enough:

```text
client-acme/
  context.md
  decisions.md
  corrections.md
  task-brief.md
  receipts/
```

The names are less important than the responsibilities.

`context.md` holds current facts. `decisions.md` records choices that affect future work. `corrections.md` stores reusable corrections. `task-brief.md` defines the present job. `receipts/` records what the agent read and produced.

This is business memory because the record survives the conversation. It can be reviewed without opening a vendor transcript. It can also be supplied to another model when the current one changes.

## Make the source order explicit

An agent should not guess which file wins when two records conflict.

Put the order in the task brief:

```text
source_order:: context.md, decisions.md, corrections.md
output:: weekly-client-update.md
must_include:: work completed, current blocker, next approved action
must_exclude:: unapproved scope, internal speculation
```

If `context.md` says the package costs `<new-amount>` and an old chat says `<old-amount>`, the brief should make the current file authoritative. If the conflict cannot be resolved from the named sources, the agent should stop and present it.

## Inspect one concrete file

Create `task-brief.md` for the recurring job. A future reviewer should be able to answer five questions from that file:

1. What result is being prepared?
2. Which files may the agent read?
3. Which source wins during a conflict?
4. What action requires approval?
5. What receipt proves the run?

Then run a cold-start test. Open a new conversation or agent session. Provide only the named files and the task. Do not add remembered context during the test.

Pass means the draft uses the current facts, respects exclusions, cites its source files, and stops at the stated boundary. Fail means you update the record or retrieval path, not improvise a longer opening prompt.

## What week one looks like

The first step is a reconstruction inventory. Write down what you supplied manually during the last run: account identity, current scope, active decision, required source, known correction, output format, and approval owner. Mark which items already exist in maintained records and which live only in memory or chat history.

The second step is source confirmation. Open each proposed source and ask who owns it, when it was verified, and whether it applies to the selected job. Do not copy every nearby document. The goal is a small packet that can support one output.

The third step is the cold run described above, with only the packet and task brief. Record every missing fact, conflict, and ambiguous instruction.

The fourth step is correction. Update the controlling file, not the chat. Run the same test again and write a receipt that says what changed.

A week-one checklist can look like this:

```text
selected_job:: weekly client status brief
repeated_explanations:: client identity, current scope, active blocker, approved next action
confirmed_sources:: client-context.md, decisions.md, work-log.md
missing_source:: current approval owner
cold_run_status:: blocked
correction:: approval owner added to task-brief.md
rerun_status:: prepared for review
external_actions:: none
```

Week one proves whether the reconstruction can be replaced by records you control.

"Week one" describes the first bounded work period, not a universal delivery guarantee. Source access, buyer review, data sensitivity, and unresolved conflicts can change the schedule. The finish line remains the tested path and recorded disposition.

## Know what changed: persistent business context for AI

The improvement is not that the agent "remembers" more. The improvement is that the business can inspect what the agent receives.

Before, a person assembled context from memory and hoped the model interpreted it correctly. After, the job starts from a named packet, reports missing inputs, and ends at a written boundary. A different person can review the same sources. A correction can change the next run.

This also sets a realistic limit. A source record does not eliminate judgment. It does not guarantee a correct draft. It makes the inputs, rules, and misses available for review.

A record fixes context gaps, not knowledge gaps. If the model does not know the subject, the record does not teach it. A vague prompt such as "make this better" also stays vague with a record behind it, so fix the prompt first.

You may not need the record at all. One-off questions with no business context lose nothing to the reset. A single simple job, such as email drafts, can fit in custom instructions. If a Project or custom GPT already serves the job, keep it. If your processes are not defined yet, define them before you write the record.

## Approval and stopping boundary: persistent business context for AI

The system must stop when a required fact is missing, two authoritative records conflict, or the requested action would send, publish, book, spend, delete, or change a live system.

The agent may prepare the exact proposed action and its supporting source. It must ask for approval before execution.

## Keep the files

The [Obsidian business-memory guide](/articles/obsidian-ai-memory-for-business) shows one way to hold the source record, retrieval method, correction path, approval rule, and run receipt outside the vendor account.


## Questions about Stop Re-Explaining Your Business to AI

### Got tired of pasting the same context into free ChatGPT every session, anyone has the same problem?

Start with one recurring job and create a small folder containing current context, decisions, corrections, a task brief, and receipts. The agent reads that maintained record instead of asking the operator to reconstruct the business from memory and old chats.

### How can I stop repeating my business context in every new chat?

Separate source facts from task instructions, then make the source order explicit in the task brief. Run a cold-start test with only the named files, and update the controlling file when the test identifies a missing fact.

### Do you ever have to re-explain the same context to ChatGPT because it forgets between sessions?

Use a durable path with context.md, decisions.md, corrections.md, task-brief.md, and receipts. The path should identify current facts, approved choices, reusable corrections, output requirements, and the boundary where the agent must stop.

## Save the visual summary of persistent business context for AI

Reconstruction is hidden work | Give the agent a path, not a pile | Make the source order explicit | Keep the files

[Download the PNG](/infographics/stop-re-explaining-your-business-to-ai-1200x1500.png)


## More reading on persistent business context for AI

- [Types of Memory Used in AI Systems and Which Ones a Business Must Own.](/articles/types-of-memory-ai-systems-business)
- [Compare a DIY Obsidian and Claude Code setup with a scoped build and choose your route](/articles/diy-obsidian-claude-code-vs-scoped-build)
- [Why Plain Text Is the Durable Layer for AI Memory](/articles/plain-text-ai-memory)

----

```text
                  .|########||.                                       .|########||.                                       .|########||.
               |##||.      .||##|.                                 |##||.      .||##|.                                 |##||.      .||##|.
             |#|.              .|#|.                             |#|.              .|#|.                             |#|.              .|#|.
           |#|                    |#|                          |#|                    |#|                          |#|                    |#|
         .#|                        |#.                      .#|                        |#.                      .#|                        |#.
        .#.                          .#|                    .#.                          .#|                    .#.                          .#|
       |#.                            .#|                  |#.                            .#|                  |#.                            .#|
      |#             ......             #|                |#             ......             #|                |#             ......             #|
     .#           ||#########|           #|              .#           ||#########|           #|              .#           ||#########|           #|
    .#.         |######||######|.        .#.            .#.         |######||######|.        .#.            .#.         |######||######|.        .#.
    #.        .##|###|##|#|######|        .#            #.        .##|###|##|#|######|        .#            #.        .##|###|##|#|######|        .#
   ||        |##|#||||||||||||#||#|        ||          ||        |##|#||||||||||||#||#|        ||          ||        |##|#||||||||||||#||#|        ||
   #        |#||||||||||||||||||||#|        #.         #        |#||||||||||||||||||||#|        #.         #        |#||||||||||||||||||||#|        #.
  ||       |#||||||||||||||||||||||#|       ||        ||       |#||||||||||||||||||||||#|       ||        ||       |#||||||||||||||||||||||#|       ||
  #       .#||||||||||||||||||||||||#|       #        #       .#||||||||||||||||||||||||#|       #        #       .#||||||||||||||||||||||||#|       #
 ||   ....|||#||||||##|#|||#|#||##||||....|. ||      ||   ....|||#||||||##|#|||#|#||##||||....|. ||      ||   ....|||#||||||##|#|||#|#||##||||....|. ||
 #.  .  ....|#  ....#|||   ||| .#|||  ....#. .#      #.  .  ....|#  ....#|||   ||| .#|||  ....#. .#      #.  .  ....|#  ....#|||   ||| .#|||  ....#. .#
 #   .  ||||#| .#####||  . .#| .#|||  ||||#   #.     #   .  ||||#| .#####||  . .#| .#|||  ||||#   #.     #   .  ||||#| .#####||  . .#| .#|||  ||||#   #.
.|   |||||  #. |#|||||. ||  #. |#||. .|||||   ||    .|   |||||  #. |#|||||. ||  #. |#||. .|||||   ||    .|   |||||  #. |#|||||. ||  #. |#||. .|||||   ||
||   |....  #. ....|#.      |. ...|. ....||   ||    ||   |....  #. ....|#.      |. ...|. ....||   ||    ||   |....  #. ....|#.      |. ...|. ....||   ||
#.  .||||||##||||||##||####|||||||#||||||#|   .#    #.  .||||||##||||||##||####|||||||#||||||#|   .#    #.  .||||||##||||||##||####|||||||#||||||#|   .#
#    .||####|###########################|.     #    #    .||####|###########################|.     #    #    .||####|###########################|.     #
#      .#||||||.#.|| # |. ..# #| #|||||#|      #    #      .#||||||.#.|| # |. ..# #| #|||||#|      #    #      .#||||||.#.|| # |. ..# #| #|||||#|      #
#      .#|||||| ..  |# ## |#| ...#|#|||#|      #    #      .#|||||| ..  |# ## |#| ...#|#|||#|      #    #      .#|||||| ..  |# ## |#| ...#|#|||#|      #
#      .#|||#|# .# .#| #| ##|.#..#|||||#|      #    #      .#|||#|# .# .#| #| ##|.#..#|||||#|      #    #      .#|||#|# .# .#| #| ##|.#..#|||||#|      #
#      .#||||||############|######|||||#|      #    #      .#||||||############|######|||||#|      #    #      .#||||||############|######|||||#|      #
#   |...||#||||||#||||#|#||||||#|||||#||| |.|  #    #   |...||#||||||#||||#|#||||||#|||||#||| |.|  #    #   |...||#||||||#||||#|#||||||#|||||#||| |.|  #
#. .. ||||# .|||##|.  |#| .|| || .|||#. #|  # .#    #. .. ||||# .|||##|.  |#| .|| || .|||#. #|  # .#    #. .. ||||# .|||##|.  |#| .|| || .|||#. #|  # .#
|| |  ...||  ...##| |  #| .|. |. #####  .. .| ||    || |  ...||  ...##| |  #| .|. |. #####  .. .| ||    || |  ...||  ...##| |  #| .|. |. #####  .. .| ||
|| ||||| || ||||#|  .  |. |. |#. ||||| |#| || ||    || ||||| || ||||#|  .  |. |. |#. ||||| |#| || ||    || ||||| || ||||#|  .  |. |. |#. ||||| |#| || ||
.# |.....#|....|#.||||.|||##.|#|....||.#||.#. #.    .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #.    .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #.
 #.|||||||#############################| ||| .#      #.|||||||#############################| ||| .#      #.|||||||#############################| ||| .#
 ||       ##||#||#||||||#||#||#||#||##.      ||      ||       ##||#||#||||||#||#||#||#||##.      ||      ||       ##||#||#||||||#||#||#||#||##.      ||
  #       .#|||||||||#||#|||||||||||#|       #        #       .#|||||||||#||#|||||||||||#|       #        #       .#|||||||||#||#|||||||||||#|       #
  ||       |#||||||||||||||||||||||#|       ||        ||       |#||||||||||||||||||||||#|       ||        ||       |#||||||||||||||||||||||#|       ||
  .#        |#||||||||||||||||||||#|        #.        .#        |#||||||||||||||||||||#|        #.        .#        |#||||||||||||||||||||#|        #.
   ||        |#||||||||||||||||||#|        ||          ||        |#||||||||||||||||||#|        ||          ||        |#||||||||||||||||||#|        ||
    #.        .######|#|#########|        .#            #.        .######|#|#########|        .#            #.        .######|#|#########|        .#
    .#.         |#####||||#####|         .#.            .#.         |#####||||#####|         .#.            .#.         |#####||||#####|         .#.
     |#           ||########||           #.              |#           ||########||           #.              |#           ||########||           #.
      |#             ......             #|                |#             ......             #|                |#             ......             #|
       |#.                            .#|                  |#.                            .#|                  |#.                            .#|
        |#.                          .#.                    |#.                          .#.                    |#.                          .#.
         .#|                        |#.                      .#|                        |#.                      .#|                        |#.
           |#|                    |#|                          |#|                    |#|                          |#|                    |#|
            .|#|.              .|#|.                            .|#|.              .|#|.                            .|#|.              .|#|.
               |##||.      .||##|                                  |##||.      .||##|                                  |##||.      .||##|
                 .||########||.                                      .||########||.                                      .||########||.

Scale With Search  2026  [scalewithsearch.com](https://scalewithsearch.com)
```
