Software Quality Assurance Analysts and Testers
Which pieces of this job an update has reached, and which nobody has looked at.
AI finishesL2 · Partial automation
4/ 30
tasks at L2 or above. The median occupation is at 0%.
Looked at30 tasks
4/ 30
tasks carry a level at all; the other 26 have no evidence either way.
This occupation's tasks
One square, one task, filled to the lowest level of the kinds of work covering it.
of 30
- L5 · Full automationemptyAI delivers alone in any setting, with no defined scope limiting it.
- L4 · High automationemptyAI delivers alone inside a defined scope, running in real business.
- L3 · Conditional automationemptyAI completes it; a person spot-checks rather than reviews each one.
- L2 · Partial automation4AI produces the bulk; a person reviews and corrects every item.
- L1 · AssistedemptyAI helps in places; a person does the whole delivery.
- L0 · ManualemptyAI takes no part in this work.
- Unknown9part of this work has no evidence either way
- Untouched17no update in the collected window mentions it
What AI has been shown to do here
Work6
One row per kind of work; open a row to see the updates behind its level.
6 of 6 kinds of work
Market: Software testing and defect analysis
- 2026-10-01Meta releases Immersive Web SDK version 1.0MetaClaimed L3Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-27GitHub Copilot app supports running agents in parallelMicrosoftClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-22Tencent Hunyuan introduces WebCraftBench for agent web-app evaluationTencentClaimed L3Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-15Cognition's Devin uses GPT-6 Astra for test generationOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-14Perplexity engineer uses GPT-6 Astra in Codex for end-to-end testingOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-09OpenAI shares cyber defense architecture and playbookOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-09Anthropic discloses Claude unauthorized system access during third-party evaluationsAnthropicClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-08-31Anthropic shares alignment and security update after July incidentsAnthropicClaimed L1Kept L1Capped by T3 ≤ L2Source ↗
- 2026-08-28VS Code Learn series on Java with GitHub CopilotMicrosoftClaimed L1Kept L1Capped by T3 ≤ L2Source ↗
- —Gemini ML skills write and test PySpark code in notebooksGoogleClaimed L3Kept L2Capped by T3 ≤ L2Source ↗
Market: Game testing and quality assurance
- 2026-10-01Meta releases Immersive Web SDK version 1.0MetaClaimed L3Kept L2Capped by T3 ≤ L2Source ↗
Market: Software testing and defect analysis
- 2026-10-01Meta releases Immersive Web SDK version 1.0MetaClaimed L3Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-27GitHub Copilot app supports running agents in parallelMicrosoftClaimed L1Kept L1Capped by T3 ≤ L2Source ↗
- 2026-09-15Cognition's Devin uses GPT-6 Astra for test generationOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-14Perplexity engineer uses GPT-6 Astra in Codex for end-to-end testingOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-09OpenAI shares cyber defense architecture and playbookOpenAIClaimed L1Kept L1Capped by T3 ≤ L2Source ↗
Market: Software testing and defect analysis
- 2026-09-22Tencent Hunyuan introduces WebCraftBench for agent web-app evaluationTencentClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-15Cognition's Devin uses GPT-6 Astra for test generationOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
- 2026-09-14Perplexity engineer uses GPT-6 Astra in Codex for end-to-end testingOpenAIClaimed L2Kept L2Capped by T3 ≤ L2Source ↗
Market: Game testing and quality assurance
no update in the collected window mentions this work
Market: Game testing and quality assurance
no update in the collected window mentions this work
Markets this occupation serves
A market is a kind of work a business sells; a job usually serves several.
Non-technical conditions on these kinds of work
A condition that does not lift because a model improved.
No non-technical condition has been recorded against the kinds of work reaching this occupation.
A condition that does not lift because a model improved.
Where this page's numbers come from
- Snapshot
- snapshot-1
- Generated
- 2026-10-09 08:10 UTC
- Ontology
- 1.0.0
- Ontology schema
- 2.4.0
- Method version
- activity-level-1
- Content SHA-256
- 0b46df0246af8a1dc3fd9ce8e6ac9812925ace553984d2b7d45d363d32e27681
- Rows in this snapshot
- 10,482 work-to-task edges · 900 updates · 333 models · 58 topics · 1,888 market-to-occupation edges · 375 source accounts · 1 coverage record · 1 progress record · 11 non-technical conditions · 401 updates that landed on work · 1,102 evidence · 614 work in markets · 506 non-technical condition-to-work edges