--- title: "AI agents for research: 36 real examples" canonical: "https://agenticfeats.com/categories/research/" receipts: 36 shown: 18 page: 1 pages: 1 --- > From Agentic Feats: sourced examples of AI agents at work. Map of the site for agents: https://agenticfeats.com/llms.txt # AI agents for research: 36 real examples Agentic Feats lists 36 sourced examples of AI agents doing research work; 18 link the artifact; most common agent: Claude Code (6). Newest first. ## Feats, newest first - [Restaurant search with computer use](https://agenticfeats.com/c/restaurant-search-with-computer-use-claude-computer-use.md) (Claude computer use; Shown; 2026-10-13): Claude's computer-use agent searched Google for vegetarian restaurants near Bond Street and saved results to a notepad file on a remote machine. Result: It created the text file as asked, but after clicking a result it took a screenshot too early and hallucinated the answer. - [Hands-on Operator test](https://agenticfeats.com/c/hands-on-operator-test.md) (Operator; Numbers; 2026-10-13): Casey Newton tested OpenAI's Operator on walking tours, a Gatsby lesson plan, Instacart groceries, and an Uber price check. Result: Operator handled tour search and lesson plan but was slow; adding three grocery items took 15 minutes and an Uber price check took five. - [NSF project summary drafting](https://agenticfeats.com/c/nsf-project-summary-drafting-cursor.md) (Cursor; Claim; 2026-10-13): The author used Cursor in Agent mode with Claude 3.7 Sonnet and two rules files to draft an NSF Project Summary from two papers. Result: The draft was usable but weaker than accepted NSF summaries; thinking mode helped, though it began hallucinating and misreading the sources. - [School financials from ProPublica](https://agenticfeats.com/c/school-financials-from-propublica-operator.md) (Operator; Numbers; 2026-10-13): Operator searched ProPublica Nonprofit Explorer for independent schools and logged their financial data into a Google Sheet; o1 then built a dashboard. Result: Operator collected data for about 25 of 100 schools before the author stopped it; o1 then built a working chart dashboard from that data. - [Operator concert and product research](https://agenticfeats.com/c/operator-concert-and-product-research.md) (Operator; Shown; 2026-10-13): Operator, OpenAI's browser agent, was tried on finding Berlin classical concerts and researching Anki competitors, using prompts and screen recordings. Result: Both tasks failed. The concert search returned three wrong concerts after 33 minutes, and the Anki competitor list had incorrect data. - [Manus tested on research tasks](https://agenticfeats.com/c/manus-tested-on-research-tasks.md) (Manus; Numbers; 2026-10-13): Manus researched China tech reporters, NYC two-bedroom listings, and Innovators Under 35 candidates for an MIT Technology Review reporter, who steered it with feedback. Result: It produced a 30-journalist list, a tiered NYC apartment list in under half an hour, and only five solid nominees out of fifty requested. - [Synthetic AGI attitudes survey study](https://agenticfeats.com/c/synthetic-agi-attitudes-survey-study-manus.md) (Manus; Shown; 2026-10-13): Manus planned a survey study on student attitudes to AGI, wrote a Google Apps Script for the form, simulated 1,000 students, and analyzed the data. Result: From one short prompt, Manus produced a publishable-style report with charts from 1,000 synthetic responses. The data is made up. - [Operator review: tweets, news, builders](https://agenticfeats.com/c/operator-review-tweets-news-builders.md) (Operator; Shown; 2026-10-13): The author spent $200 on ChatGPT Operator and tested it on tweet analysis, AI news summaries, website builder pricing lists, and article summaries. Result: It returned duplicate and months-old tweets, summarized one news source, copied a single listicle, and stopped to ask permission mid-task. The author says productivity dropped. - [Manus tests on hotel booking engines](https://agenticfeats.com/c/manus-tests-on-hotel-booking-engines.md) (Manus; Numbers; 2026-10-13): Manus.ai was prompted to book a September stay at hotels on 11 popular booking engines to test its navigation. Result: Manus succeeded on 8 of 11 booking engines, three of those with delays, and failed on D Edge, Simple Booking and Travelanium. - [UNESCO education data spreadsheet](https://agenticfeats.com/c/unesco-education-data-spreadsheet-chatgpt-agent.md) (ChatGPT agent; Shown; 2026-10-13): ChatGPT Agent pulled UNESCO student-teacher ratio and teaching staff data and built a multi-tab spreadsheet in a single 14-minute run. Result: The Agent worked for 14 minutes on the UNESCO task; the author tested it against o3-pro and ran four other work-style tasks. - [Booking tests with ChatGPT Agent](https://agenticfeats.com/c/booking-tests-with-chatgpt-agent.md) (ChatGPT agent; Numbers; 2026-10-13): ChatGPT Agent was tested on restaurant and software demo bookings, with mixed results across three attempts. Result: The VICE booking took over eight minutes and four tries on the time picker, while the SE Ranking demo booking took 37 seconds. - [Annotated bibliography on big data](https://agenticfeats.com/c/annotated-bibliography-on-big-data-chatgpt-agent.md) (ChatGPT agent; Shown; 2026-10-13): ChatGPT Agent searched academic databases, filtered peer-reviewed open-access papers, and compiled an annotated bibliography on big data in education and healthcare decision-making. Result: It produced a review-ready draft with citations and a gaps synthesis in under ten minutes, versus half a day by hand. - [AI agents play free games](https://agenticfeats.com/c/ai-agents-play-free-games.md) (Claude Code, Other; Shown; 2026-10-13): Seven AI agents, each with a Linux computer and a group chat, tried to complete as many online games as possible in a week. Result: No agent won a game. Gemini 2.5 Pro made the most progress, reaching veteran footman rank in the idle game Progress Knight. - [Competitor analysis deck and table](https://agenticfeats.com/c/competitor-analysis-deck-and-table-chatgpt-agent.md) (ChatGPT agent; Shown; 2026-10-13): ChatGPT Agent mode researched the top 5 UK coffee subscription companies and built a competitor deck plus a spreadsheet of each company's top 5 USPs. Result: One prompt returned a PDF deck and a USP spreadsheet in 18 minutes, with positioning, pricing models and a SWOT section. - [Emailed AI news briefing](https://agenticfeats.com/c/emailed-ai-news-briefing-perplexity-comet.md) (Perplexity Comet; Shown; 2026-10-13): Perplexity Comet searched major outlets for the day's AI news, picked the top five stories, and emailed the author a brief. Result: A five-story AI news brief arrived by email from one prompt; the author says plain LLMs only draft text and cannot send it. - [AI browser shopping test](https://agenticfeats.com/c/ai-browser-shopping-test-perplexity-comet.md) (ChatGPT agent, Other, Perplexity Comet; Shown; 2026-10-13): DreamHost tested five AI browsers on one Amazon task: research grinder coffee makers under $100, tabulate five, add two to cart. Result: Perplexity Comet completed the task and stopped at the cart. ChatGPT Atlas also reached checkout but was too slow to be usable. - [App review analysis for PMs](https://agenticfeats.com/c/app-review-analysis-for-pms-cursor.md) (Cursor; Shown; 2026-10-13): Cursor analyzed a CSV of a couple thousand Walmart App Store reviews and returned takeaways with ASCII diagrams. Result: In 30 seconds it returned top complaints ranked by frequency, a severity versus frequency matrix, and a user journey map showing where frustration peaks. - [Genetic and bloodwork health plan](https://agenticfeats.com/c/genetic-and-bloodwork-health-plan-claude-code.md) (Claude Code; Claim; 2026-10-13): Claude Code analyzed the author's 23andMe raw genetic file, bloodwork, artery plaque scans, and supplement routine in a two-hour back and forth about health. Result: The author came away with a health plan tailored to their genes and current health, built from the combined datasets. - [Agent-built research knowledge bases](https://agenticfeats.com/c/agent-built-research-knowledge-bases-openclaw.md) (OpenClaw; Shown; 2026-10-13): OpenClaw learned to use a YouTube summariser, a paper extraction tool, and QMD, then built searchable Markdown knowledge bases for videos and research papers. Result: The author now has two dense, searchable knowledge bases, one from YouTube videos and one from research papers, which they say speeds up understanding. - [YouTube research into Obsidian graph](https://agenticfeats.com/c/youtube-research-into-obsidian-graph-claude-code.md) (Claude Code; Shown; 2026-10-13): Claude Code searched YouTube, loaded 20 videos into NotebookLM, asked cited questions, and saved everything to Obsidian as linked notes. Result: The author turned 20 videos into a browsable knowledge graph of sources and topics in their vault. A citation audit found about 60% strong matches. - [Manus 15-task agent test](https://agenticfeats.com/c/manus-15-task-agent-test.md) (Manus; Numbers; 2026-10-13): The JustPickAi editorial team gave Manus 15 real-world tasks across research, data, coding and content, then timed and scored each one. Result: In the published table, most tasks succeeded in 4 to 15 minutes, with 70% full success, 20% partial and 10% failure. - [Weekly Substack ranking report](https://agenticfeats.com/c/weekly-substack-ranking-report-claude-in-chrome.md) (Claude Cowork, Claude in Chrome; Shown; 2026-10-13): Claude in Chrome scrapes the top 250 Substack newsletters from Reletter each week, then Claude Cowork turns the data into a gains and losses report. Result: The author's weekly chart scraping and rank analysis now runs as two scheduled tasks every Tuesday, replacing manual copy-paste and taking a couple of minutes. - [Conference attendee research briefing](https://agenticfeats.com/c/conference-attendee-research-briefing-claude-code.md) (Claude Code; Numbers; 2026-10-13): Claude Code researched 40 conference attendees in parallel through Chrome MCP, using the author's LinkedIn login, and compiled everything into one briefing. Result: A full participant briefing on 40 people was ready for review within 5 minutes, replacing an evening of manual LinkedIn clicking. - [X feed research for posts](https://agenticfeats.com/c/x-feed-research-for-posts-claude-computer-use.md) (Claude computer use; Shown; 2026-10-13): Claude computer use, sent from the phone via Dispatch, opened Chrome, read the author's logged-in X feed, and picked posts worth writing about. Result: In about 90 seconds it returned four posts with view and like counts, plus a suggested angle for each. - [Adyen equity research in Cowork](https://agenticfeats.com/c/adyen-equity-research-in-cowork-claude-cowork.md) (Claude Cowork; Shown; 2026-10-13): Claude Cowork pulled Adyen filings and data, built an Excel model, ran a DCF, wrote an investment memo, and set up a weekly update task. Result: A full equity research workflow on Adyen, from data pull to investment memo, was done in under 15 minutes using prompts only. - [Daily intelligence briefing to Slack](https://agenticfeats.com/c/daily-intelligence-briefing-to-slack-claude-cowork.md) (Claude Cowork; Numbers; 2026-10-13): Claude Cowork runs a scheduled task each morning that searches research pillars, dedupes against a Notion list, and posts a curated 10-story briefing to Slack. Result: Raghav Mehra cut morning source-checking from 45+ minutes to about 5 minutes of reading a pre-curated briefing. - [Weekly competitive briefing skill](https://agenticfeats.com/c/weekly-competitive-briefing-skill-hermes-agent.md) (Hermes Agent; Numbers; 2026-10-13): Hermes Agent ran the same Monday competitive monitoring prompt for six weeks and rewrote its own underlying skill four times. Result: The same task fell from 20 minutes in week one to 12 in week four and 8 by week six, with better competitor patterns surfaced. - [Agent researches its own competitors](https://agenticfeats.com/c/agent-researches-its-own-competitors-custom-agent.md) (Custom agent; Numbers; 2026-10-13): Daemora, a self-hosted agent, searched Reddit, HN, Twitter and review sites unprompted and wrote a competitor analysis and go-to-market plan from one task. Result: It wrote a 3,000 word document with competitor pain points, 40 named target users, and an outbound playbook. - [Client notes to Belmond shortlist](https://agenticfeats.com/c/client-notes-to-belmond-shortlist-codex.md) (Codex; Numbers; 2026-10-13): Codex Desktop read a client notes Word doc, checked Belmond's live booking site, and built a three-option comparison matching the client's past trip preferences. Result: Codex returned a ranked comparison of three Belmond properties with availability and nightly rate ranges for a sample window, plus a more detailed Word doc. - [MRI second opinion](https://agenticfeats.com/c/mri-second-opinion-claude-code.md) (Claude Code; Shown; 2026-10-13): Claude Code with Opus 4.8 read a DICOM shoulder MRI export, then ran a subagent-based arbitration against the human radiology report. Result: Opus found no tear where the doctor reported a Grade III partial-thickness tear. The author is unsure which reading to trust. - [Research workflow automation with Codex](https://agenticfeats.com/c/research-workflow-automation-with-codex.md) (Codex; Claim; 2026-10-13): Codex runs a weekly arXiv search, triages new papers, feeds a Zotero library, and turns hypotheses into cheap experiment tests for an AI safety researcher. Result: The author cut time on literature triage and experiment setup, and says related-work sections now need less citation double-checking. No numbers are given. - [Belgian supermarket Coca-Cola prices](https://agenticfeats.com/c/belgian-supermarket-coca-cola-prices-grok-bot.md) (Grok Bot; Numbers; 2026-10-13): Grok Bot compared Coca-Cola prices across Belgian supermarket sites, then handed the browser to the author to clear a Carrefour human check. Result: It returned a price table and promotions, noted Aldi and Lidl gaps, and revised the best price per liter to Carrefour's 1,48 euro two-liter. - [ChatGPT Dots day-one test](https://agenticfeats.com/c/chatgpt-dots-day-one-test.md) (Other; Shown; 2026-10-13): Julian Goldie tested OpenAI's new ChatGPT Dots agent on launch day with research, a website build, computer use, and a restaurant booking. Result: Research, site build and restaurant booking worked, but computer use was very slow and the call button failed. - [Memecoin vetting bot team](https://agenticfeats.com/c/memecoin-vetting-bot-team-grok-bot.md) (Grok Bot; Numbers; 2026-10-13): Grok Bot learned the author's 4-minute memecoin check from one demo, then ran it through a scout, checker, and reporter bot team. Result: Over 30 days it checked 3,140 coins, skipped 2,980, raised 160 alerts, and the author made 52 trades for +$9,320. - [Research workflow with Claude Code](https://agenticfeats.com/c/research-workflow-with-claude-code.md) (Claude Code; Claim; 2026-10-13): Claude Code with MCP servers handles literature search, Zotero notes, Gmail alert newsletters, and codebase questions for a hydrology researcher. Result: The author reports faster literature search, Zotero notes written automatically, and accurate answers about the ATS hydrology codebase through a custom ats-mcp server. - [Meeting minutes into NotebookLM](https://agenticfeats.com/c/meeting-minutes-into-notebooklm-claude-in-chrome.md) (Claude in Chrome; Numbers; 2026-10-13): Claude in Chrome, on the Haiku model, extracted exact text of three years of W3C task force meeting minutes into markdown for NotebookLM. Result: Three years of minutes (2022-2024) were extracted in about 3 to 4 hours across three sessions and loaded as one source per year.