Real Agent Cases

Home › Category

Research

20 source-checked cases.

  • Report global NAND market share with Gemini Deep ResearchImproved promptEvidence: Screenshots or logsFailure

    Asked for a simple report on global NAND market share, Gemini Deep Research returned figures the poster called an incredible failure when compared with Sandisk's investor-day slide and Claude's answer to the same query.

    Gemini Deep Research · Research · Source date:

  • Write a one-page sales brief on a target company with ManusRecreated promptEvidence: Article or interview

    Asked for a B2B sales brief on Google for a tl;dv pitch, Manus produced a one-page brief with revenue, tech stack, key decision makers, recent announcements and a sourced section mapping Google's pain points to tl;dv features; the reviewer says it took two minutes.

    Poster says:About two minutes

    Manus · Research · Source date:

  • Working from the family's archive scans, Claude Code with memory and the Claude browser plugin cross-checked church books, ship manifests and naturalization records, built a 300-page family wiki, and found two more generations in a day; the family is now checking it for errors.

    Poster says:300-page family wiki

    Claude Code, Claude in Chrome, Perplexity Comet · Research · Source date:

  • Fact-check a viral research-paper claim with Claude in ChromeRecreated promptEvidence: Screenshots or logs

    Asked to fact-check a post claiming Bitcoin would 'surge 500%' based on a study, Claude for Chrome reported in about 2 minutes that the paper wasn't in any academic database, the 500% maths was wrong, a commenter said it was AI-generated, and the paper itself said it wasn't a prediction.

    Poster says:About 2 minutes

    Claude in Chrome · Research · Source date:

  • Working for over 45 minutes, GPT-5 Pro found correct FEINs for 4 of 6 subsidiaries on the Florida Division of Corporations site and one address via a FINRA BrokerCheck report; the poster double-checked the FEINs.

    GPT-5 Pro · Research · Source date:

  • Summarize English and Tamil YouTube videos with CometImproved promptEvidence: Described by posterPartial

    Comet Assistant summarized English YouTube conversations quickly and well, but on a Tamil discussion it had to play the whole video and made a few mistakes, missing some crucial points.

    Perplexity Comet Assistant · Research · Source date:

  • Test five web tasks with Project MarinerImproved promptEvidence: Shown on videoPartial

    In a test of 5 tasks, Project Mariner found YouTube video view counts, but failed to generate relevant video ideas, couldn't preview HTML properly, refused to log into email and couldn't send emails.

    Project Mariner · Research · Source date:

  • Find links to content with Project Mariner (no links returned)Improved promptEvidence: Described by posterFailure

    Asked to find links to content, Project Mariner found the content but wouldn't return the links, which the poster said made the result not very helpful.

    Project Mariner · Research · Source date:

  • Buying a new house, Dan Fitzpatrick ran Gemini Deep Research and ChatGPT Deep Research on family life in the area he was moving to; he says Gemini 'won this by miles' with a wider range of research, more questioning of sources and a more thorough report, delivered as a Google Doc.

    Gemini Deep Research · Research · Source date:

  • Complete a Qualtrics survey with OpenAI OperatorRecreated promptEvidence: Screenshots or logs

    The poster, a survey researcher, asked Operator to complete a Qualtrics survey they had created; they call the result very promising for Operator but very concerning for survey researchers.

    OpenAI Operator · Research · Source date:

  • Indian Express's reviewer asked Grok 3 DeepSearch about the state of the house sparrow in India; he reports it scoured 101 sources in 46 seconds before presenting its findings.

    Poster says:101 sources in 46 seconds

    Grok DeepSearch · Research · Source date:

  • Connected read-only to Gmail, Claude Cowork found 12 relevant PR pitches in about a minute, then narrowed to eight on-record quotes with publish permission that the journalist verified before use.

    Claude Cowork · Research · Source date:

  • Asked only about cmarkgfm, Claude Code found six other Python Markdown libraries by itself, benchmarked all seven and wrote a report with charts, in which cmarkgfm came out on top; the author notes he hasn't been reviewing these reports in great detail.

    Claude Code · Research · Source date:

  • Compare top social media tools in a spreadsheet with ChatGPT agent and ClaudeRecreated promptEvidence: Screenshots or logsBeginner-friendly

    Allie K. Miller asked both ChatGPT Agent Mode and Claude's Excel feature to compare top social media tools, and Claude finished in 3 minutes vs 8 with denser decision support, though some prices mixed annual vs monthly.

    ChatGPT agent, Claude (Excel) · Research · Source date:

  • Using scraped HMRC manuals and legislation plus a CLAUDE.md that launches parallel subagents, Claude Code answered an ATT exam question in about 5 minutes, earning roughly 2.5 of 3 marks by the author's grading.

    Claude Code · Research · Source date:

  • Build an Nvidia DCF valuation spreadsheet with ChatGPT agentRecreated promptEvidence: Screenshots or logsBeginner-friendly

    ChatGPT agent ran for 37 minutes, walked through DCF assumptions and returned a downloadable Excel file with formulas and numbers.

    ChatGPT agent · Research · Source date:

  • Build an early-retirement (FIRE) plan with ChatGPT agentRecreated promptEvidence: Shown on videoBeginner-friendly

    With early access, the poster had ChatGPT agent spend 20 minutes researching Vancouver tax law, modeling savings and FIRE scenarios, and producing a downloadable presentation.

    ChatGPT agent · Research · Source date:

  • Find and summarize files in Google Drive with Comet AssistantImproved promptEvidence: Shown on videoPartial

    Comet Assistant kept trying until it found the right files in Google Drive and found a way around opening them, but its summary skipped parts of a PRD after reading only about half.

    Perplexity Comet Assistant · Research · Source date:

  • Compile a list of China-tech reporters with ManusRecreated promptEvidence: Article or interviewBeginner-friendly

    After being pushed to be thorough, Manus produced a downloadable list of 30 journalists covering China tech with their outlets and notable work, though paywalls and captchas got in its way.

    Manus · Research · Source date:

  • Create an interactive elevator-pitch course with ManusEvidence: Article or interviewBeginner-friendly

    Manus researched the topic and built a complete multi-module course website on elevator pitching; a second prompt added interactive quizzes and videos.

    Manus · Research · Source date: