Skip to content
On this page

Should you price coding tasks separately from your engineering archive?

Start with the sample a buyer needs. A coding task has to run and be graded. An engineering archive has to show how a reported problem became a fix. One company had material for both, and the same code appeared in both scopes.

Suppose a buyer asks for the history behind your software. Should you spend the week making old bugs reproducible, or matching support tickets to pull requests? Ask what they want to do with the first sample. A team evaluating a coding agent needs a bug it can reproduce and a test it can grade. A team studying engineering decisions needs a trail it can follow across tools. Preparing the wrong sample can use the whole week without answering the buyer’s question.

Before you ask for a written price, give the buyer an inventory that matches the requested sample. A corrected inventory can change what you ask the buyer to price. It cannot tell you what they will pay. Two public Rightmark listings from one software company show the distinction in dollars: one prices tasks that pass a grading check, the other prices a license to code and selected records. Both figures are asking prices.

A fix that changes a test gives you a place to start

The task listing covers 13 repositories selected for task building. In those repositories, 12,889 commits were classified as bug fixes. Of those, 2,310 changed a test in the same commit. More than half of those candidates came from the main application. If you own a similar portfolio, testing whether that application builds in a clean environment is a sensible first use of preparation time.

Where the 2,310 test-changing fixes came from
Main application1,505 (65.2%)
Other 12 repositories805 (34.8%)

One owner’s 13 selected repositories. The full width represents all 2,310 candidates. Source: SE-C2, checked November 22, 2026.

Values and scope
Where the test-changing fixes came fromCandidatesShare of the 2,310
Main application1,50565.2%
Other 12 repositories combined80534.8%
Total across the 13 repositories2,310100%

The owner’s repository check, as reported on the public task listing on November 22. 1,505 + 805 = 2,310. Dividing 2,310 by 12,889 gives 17.9% of classified fixes. The two table shares use 2,310 as their denominator. The listing gives no count of verified tasks.

Take one candidate. Isolate the regression test and run it against the code before and after the fix. Save both results: a failure on the earlier revision and a pass on the fixed one. Record the dependencies needed to repeat both runs. Three of the 13 repositories need private packages to build. Discovering that in the first sample gives the owner a choice: clear the dependency or pick another fix before promising a batch. Task listing SE-C2

A buyer licensing the engineering trail needs different proof

The second listing covers the full 138-repository code collection with the issue tracker, selected engineering chat and support-desk bug reports. It lists 58,938 merged pull requests. A buyer can follow a customer report through to a fix only where the records contain the links. The listing leaves the share of pull requests linked to an issue for measurement before contract. It also excludes private chat channels, direct messages and support conversations unrelated to bugs.

For this sample, start at the support desk. Pick a defined period and a selection rule you can describe. Count the eligible bug reports, then count how many still lead to an issue, a merged change and a fix after de-identification. Keep broken links in the denominator. Show one de-identified example and report the numerator and denominator together. A buyer can then judge both what a complete trail looks like and how often it survives in that sample. Engineering-records listing SE-C3

What the two asking prices buy

Values and scope
Non-exclusive scopePublic dollar range (USD)Unit and evidence
Verified coding task, SE-C2$1,100 to $2,000Per verified task. The listing requires a clean-container check. Low-confidence indicative ask.
Code and defined engineering records, SE-C3$130,000 to $163,000Per license for the described package. Link coverage, redaction and scope still need confirmation. Low-confidence indicative ask.

If 50 tasks qualified at the posted non-exclusive ask, the gross illustration would be 50 × $1,100 to 50 × $2,000 = $55,000 to $100,000 before preparation costs or fees. Fifty is the assumption in this illustration. Actual billable volume depends on how many tasks qualify and the terms a buyer accepts. Multiplying the 2,310 candidates by a price would skip the work that determines whether any individual task qualifies.

The broader Rightmark Price Index reports $200 to $2,000 per prepared task, with medium confidence, citing Epoch AI in January 2026. Use that for context. The task listing has its own qualification test and its own ask. The archive listing prices access to a different deliverable. Subtracting one range from the other cannot tell you what adding tickets is worth. Q4 Price Index · Public lot asks

Put shared code on both scope sheets

The public cards identify the 13 task-building repositories as part of the full 138. Counting the listings as 151 repositories would count those 13 twice. The records package has another condition: the catalog offers it on its own only if the broader company-records lot is declined or the owner splits the scope in writing. Two price sheets do not by themselves settle which rights can be sold together.

Give each scope sheet a row for the code they share. On the task sheet, identify the version, grading test and dependencies for the example you can reproduce. On the records sheet, name the included systems, date range, exclusions and measured links. Across both, specify whether the owner can still create tasks or benchmarks from the same code if one license closes. If a buyer wants exclusivity, have the permitted uses, material and duration reconciled in writing before either package is promised.

Prepare the scope you can demonstrate first. If both samples hold up and the rights can be separated, ask the buyer to price each deliverable in writing. If only one is ready, you can ask a precise question about that one without treating the rest of the archive as worthless.

Sources and scope

Rightmark checked the public listings, Q4 Price Index, Wanted board, buyer correspondence and seller intake records on November 22, 2026. The public counts describe one owner’s listed inventory. Buyer and seller records informed the preparation advice. All inventory counts and dollar figures shown here come from public sources. No independent task run, cross-system linkage measurement or completed transaction was available for this analysis. Earlier note on repository and task units · Earlier note on code, tickets and chat

The data we are looking forCatalog checked

Private software with a reproducible fix or a traceable engineering history

Rightmark’s catalog shows SE-C2 and SE-C3. The public Wanted board showed no open request at the check.

For a task, prepare a runnable before-and-after example. For an archive, prepare a de-identified bug report linked to an issue and merged fix. Identify what you own and what must be excluded.

Indicative task ask

$1,100 to $2,000

USD per verified task under SE-C2’s non-exclusive terms.

Low confidence. This is a public asking range before fees or preparation costs. It does not establish an offer for your code or proceeds to the owner.

Source: Rightmark’s SE-C2 listing, checked November 22, 2026. The SE-C3 archive has a separate per-license ask.

Start with one sample.

Describe the sample you can demonstrate and list any shared code, missing dependencies or restricted records. An appraisal can help define the scope to take to a buyer.

Keep private code and raw support conversations out of the initial description. A short inventory and the checks you have completed are enough to start.

Find out what your data could be worth.

Start with an indicative appraisal. No files or system access needed.

Appraise your data

What brings you to Rightmark?