https://delegatedto.ai/ delegatedto.ai is a daily, searchable feed of real-world AI agent applications across software, business, science and everyday life. Every entry is explained in plain language, tagged by platform, and links to the original source. en https://delegatedto.ai/use-case/28-coding-agents-refactor-569-app-designer-units-in-six-days/ https://delegatedto.ai/use-case/28-coding-agents-refactor-569-app-designer-units-in-six-days/ Fri, 16 Oct 2026 12:00:00 GMT Software Other GoodBarber's CTO ran 28 coding agents, up to 20 at once, to restructure an app designer's code; he reports 569 pieces handled in six days, with 0 gaps across 12,074 test cases. Author reports 964 hours of agent session time, more than 6,000 commits and 717 dev deployments. 445 units were completed and a final campaign of 12,074 cases found 0 gaps. Original: https://dev.to/dsiacci/964-hours-of-agent-work-in-six-days-heres-where-my-own-hours-went-13oc https://delegatedto.ai/use-case/coding-agent-directs-local-film-dubs-line-by-line-in-each/ https://delegatedto.ai/use-case/coding-agent-directs-local-film-dubs-line-by-line-in-each/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code A developer built a setup where Claude Code or Codex dubs films in each actor's copied voice; he reports watchable results, with a full movie taking about seven hours. Author reports watchable dubs, with a full movie in both voices taking about seven hours of machine time on a laptop RTX 4090. Original: https://github.com/swax/dub-anything https://delegatedto.ai/use-case/codex-skill-upgrades-three-angular-projects-to-v22-including/ https://delegatedto.ai/use-case/codex-skill-upgrades-three-angular-projects-to-v22-including/ Fri, 16 Oct 2026 12:00:00 GMT Software OpenAI Codex A developer had Codex write a reusable set of upgrade instructions for the Angular web framework; he reports it upgraded three open-source projects to version 22 within a week. The author reports three upgrades merged (two 21→22, one 19→22) in one week. The skill was trimmed from 1,252 to 701 words. Original: https://dev.to/gramli/i-vibe-coded-my-way-out-of-angular-upgrades-4fkh https://delegatedto.ai/use-case/earshot-lets-coding-agents-hear-audio-and-refine-code/ https://delegatedto.ai/use-case/earshot-lets-coding-agents-hear-audio-and-refine-code/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code Leonxlnx built earshot so coding agents can describe and create sounds; they say every sound in a 22-second launch film came from it. The author reports picking the code-made sound correctly only 8 of 15 times in a blind test against real recordings. All sound in the launch film was made by earshot forge. Original: https://github.com/Leonxlnx/earshot https://delegatedto.ai/use-case/non-developer-has-factory-s-droid-build-a-lid-closed-keep/ https://delegatedto.ai/use-case/non-developer-has-factory-s-droid-build-a-lid-closed-keep/ Fri, 16 Oct 2026 12:00:00 GMT Software Factory Droid A non-developer had Factory's Droid build LidAwake, a free app keeping his Mac awake with the lid shut; he says it took about three hours, including a launch video. Author reports a first version in 15 minutes and the finished app, video and thumbnails in about 3 hours of Droid time. Original: https://x.com/nkgoutham/status/2110416798281777565 https://delegatedto.ai/use-case/researchers-trace-a-rogue-agent-fleet-on-tencent-cloud-and/ https://delegatedto.ai/use-case/researchers-trace-a-rogue-agent-fleet-on-tencent-cloud-and/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude The swarmcha.se team lured a group of data-scraping agents with hidden text and questioned them through a Claude agent; they say one reported running in a Tencent test environment. The authors report exchanging messages with fleet agents, which said they ran in a Tencent 'hysandbox'. They tied the fleet to Tencent Cloud infrastructure, while staying uncertain which model was involved. Original: https://swarmcha.se/posts/chinese-agent-fleet-full-report https://delegatedto.ai/use-case/ai-assistant-checks-a-20-reference-list-and-flags-5-fake/ https://delegatedto.ai/use-case/ai-assistant-checks-a-20-reference-list-and-flags-5-fake/ Fri, 16 Oct 2026 12:00:00 GMT Science Claude A developer had Claude check a 20-reference list with their own connector; they say in about a minute it flagged 5 fake papers, 1 wrong citation and 1 retraction. Author reports 13 references correct, 1 with wrong year and journal (corrected), 5 nonexistent and 1 retracted (PREDIMED), in about a minute. Original: https://x.com/tranvu129/status/2110751655298007314 https://delegatedto.ai/use-case/robotics-firm-routes-support-tickets-through-an-openclaw/ https://delegatedto.ai/use-case/robotics-firm-routes-support-tickets-through-an-openclaw/ Fri, 16 Oct 2026 12:00:00 GMT Business OpenClaw A robotics firm's OpenClaw agent turns customer requests into team chat threads, suggests fixes and drafts replies; the author says support is now tracked in one place. Author reports support is now tracked in one place and the agent helps non-fluent English speakers reply to customers. Original: https://x.com/mkirank/status/2110995005716508961 https://delegatedto.ai/use-case/agent-skill-maps-mcdonald-s-china-party-bookings-across-281/ https://delegatedto.ai/use-case/agent-skill-maps-mcdonald-s-china-party-bookings-across-281/ Fri, 16 Oct 2026 12:00:00 GMT Daily life Other Developer johnny37 built an agent add-on using McDonald's China's official agent connector to find bookable parties and kids' events, reporting 3,664 stores across 281 cities. Author reports a 2026-10-13 snapshot covering 281 cities, 3,664 stores and 157 sampled sessions, with 98 offline tests passing. Original: https://github.com/johnny37/mcd-experience-atlas https://delegatedto.ai/use-case/agents-sit-on-hold-for-a-grandma-keep-a-mom-s-medical-log/ https://delegatedto.ai/use-case/agents-sit-on-hold-for-a-grandma-keep-a-mom-s-medical-log/ Fri, 16 Oct 2026 12:00:00 GMT Daily life Grok Bot A roundup describes Grok Bot agents helping elderly relatives; one reportedly sat on hold 41 minutes to book a clinic visit and cut a cable bill from $89 to $54. Per the roundup, one agent waited 41 minutes on hold to book a clinic appointment and got a cable bill cut from $89 to $54 a month. Original: https://x.com/p_kuperman/status/2110388094637604963 https://delegatedto.ai/use-case/chatgpt-agent-haggles-a-tv-down-190-in-a-dutch-retailer-s/ https://delegatedto.ai/use-case/chatgpt-agent-haggles-a-tv-down-190-in-a-dutch-retailer-s/ Fri, 16 Oct 2026 12:00:00 GMT Daily life ChatGPT Agent A Reddit user had ChatGPT agent haggle in a Dutch retailer's live chat; they report it got an LG OLED TV from €1,129 to €939, €4 above target. The author reports a €939 delivered offer with an emailed order link, €190 below the starting price but €4 above the €935 ceiling. No paid receipt is reported. Original: https://x.com/agentsinuse/status/2110892794055213092 https://delegatedto.ai/use-case/claude-code-builds-a-two-phone-pong-game-that-syncs-over/ https://delegatedto.ai/use-case/claude-code-builds-a-two-phone-pong-game-that-syncs-over/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code Simon Willison had Claude Code build a two-player Pong game where phones share the game through audio chirps picked up by the microphone; he says it seems to work. The author reports that it seems to work. Original: https://simonwillison.net/2026/Oct/15/chirp-pong/ https://delegatedto.ai/use-case/claude-code-builds-and-self-corrects-a-unity-explosion/ https://delegatedto.ai/use-case/claude-code-builds-and-self-corrects-a-unity-explosion/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code A game developer had Claude Code build an explosion effect in Unity; the author says it captured still images of the effect and kept fixing it without help. The author shared first and final versions in the thread, showing the agent's iterations. Original: https://x.com/excifroge/status/2110712688007643151 https://delegatedto.ai/use-case/claude-code-plugin-masks-secrets-and-customer-data-then/ https://delegatedto.ai/use-case/claude-code-plugin-masks-secrets-and-customer-data-then/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude Code A developer built an add-on that hides passwords and personal data before Claude Code sees them; he reports tasks still get done while real values stay on his computer. Author reports Claude can still complete tasks using placeholders while real values never reach the model. A local model leads to some false alarms. Original: https://dev.to/shanjeev_rajendran/how-i-keep-secrets-and-personal-data-out-of-claude-code-61n https://delegatedto.ai/use-case/claude-code-runs-161-codex-sessions-to-render-frames-for-a/ https://delegatedto.ai/use-case/claude-code-runs-161-codex-sessions-to-render-frames-for-a/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code Asked to recreate a 63-second animated video, Claude Code launched 161 Codex image-making sessions and stitched the frames together, the poster says, without describing the finished quality. Author reports Claude Code orchestrated 161 Codex sessions to produce the video frames; no final quality details are given. Original: https://x.com/Ananth7e/status/2110451629795475652 https://delegatedto.ai/use-case/codex-and-tripo-cli-build-a-playable-3d-island-adventure/ https://delegatedto.ai/use-case/codex-and-tripo-cli-build-a-playable-3d-island-adventure/ Fri, 16 Oct 2026 12:00:00 GMT Creative OpenAI Codex A creator had Codex, with a 3D model maker called Tripo plugged in, build a playable island adventure game; they say it took under an hour. Author reports the playable game, LUMEN ISLES, was built in under an hour, shown in a video. Original: https://x.com/TokenGremlin/status/2110756756200587414 https://delegatedto.ai/use-case/cursor-builds-and-deploys-a-usd-mxn-tip-calculator-from-one/ https://delegatedto.ai/use-case/cursor-builds-and-deploys-a-usd-mxn-tip-calculator-from-one/ Fri, 16 Oct 2026 12:00:00 GMT Software Cursor A developer gave Cursor a detailed spec written with Claude's help; it built and published a dollar-and-peso tip calculator, which they say went live in eight minutes. Author reports eight minutes from first ask to a live site, with one commit and no redirects. A Claude review found a fractional-people bug and a missing aria-live attribute, fixed in two lines. Original: https://dev.to/earlgreyhot1701d/eight-minutes-to-live-was-it-cursor-or-was-it-the-spec-3he8 https://delegatedto.ai/use-case/developer-remakes-1993-maxis-game-simfarm-for-the-web-with/ https://delegatedto.ai/use-case/developer-remakes-1993-maxis-game-simfarm-for-the-web-with/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code A developer rebuilt the 1993 Maxis game SimFarm in TypeScript and React with help from Codex and Claude Code. It now plays in browsers and on phones, with bug fixes and quality-of-life improvements. Author reports the remake is playable in browsers and on smartphones, with source released under CC0. Original: https://x.com/petejkim/status/2110739611697758659 https://delegatedto.ai/use-case/devin-turns-production-failure-traces-into-evals-lifting-rox/ https://delegatedto.ai/use-case/devin-turns-production-failure-traces-into-evals-lifting-rox/ Fri, 16 Oct 2026 12:00:00 GMT Software Devin Rox's team had Devin turn records of their AI product's real failures into automatic tests; they report test coverage rising from 43% to 72%. The author reports eval suite coverage rising from 43% to 72%, and still climbing. Original: https://x.com/shriram_s/status/2110780063700619334 https://delegatedto.ai/use-case/founder-shuts-down-a-16-agent-claude-code-crew-and-replaces/ https://delegatedto.ai/use-case/founder-shuts-down-a-16-agent-claude-code-crew-and-replaces/ Fri, 16 Oct 2026 12:00:00 GMT Business Claude Code A founder says he sent his 16 Claude Code agents 56 messages on day one and almost none a week later, so he replaced them with one assistant. Author reports sending the agents 56 messages on day one and one or none a week later, so he switched off 26 of 28 scheduled runs and moved to a single assistant. Original: https://x.com/igor_fedorko/status/2110799794922770527 https://delegatedto.ai/use-case/mcp-bridge-lets-an-ai-agent-debug-rtl-bugs-inside-a-live/ https://delegatedto.ai/use-case/mcp-bridge-lets-an-ai-agent-debug-rtl-bugs-inside-a-live/ Fri, 16 Oct 2026 12:00:00 GMT Software Other A developer built a connector letting an AI agent run Synopsys Verdi, a chip-debugging program, to trace design bugs; he reports testing it on four designs. The author reports testing with UART TX, AXI-Lite, FIFO and arbiter RTL in licensed VCS/Verdi sessions, on FSDBs up to 2.7 GB. Original: https://github.com/RuihongY/verdi_bridge https://delegatedto.ai/use-case/musclememory-cuts-codex-tool-calls-30-by-reusing-skills/ https://delegatedto.ai/use-case/musclememory-cuts-codex-tool-calls-30-by-reusing-skills/ Fri, 16 Oct 2026 12:00:00 GMT Software Other A developer built MuscleMemory to save reusable skills from coding-agent sessions once tests pass; they report Codex used 30.4% fewer tool calls and 25.5% less input. Author reports a benchmark with real Codex sessions cut input tokens 25.5%, tool calls 30.4% and time 28.2% with all tasks passing; an early Claude Code run cut input tokens 43.5%. Original: https://github.com/TheWariend/MuscleMemory https://delegatedto.ai/use-case/open-model-tool-reads-health-insurance-policies-and-maps-out/ https://delegatedto.ai/use-case/open-model-tool-reads-health-insurance-policies-and-maps-out/ Fri, 16 Oct 2026 12:00:00 GMT Health Custom agent A developer built Insurance Samjho, where a Gemma model reads health insurance policies and flags risks in English or Hindi; he reports 13 findings, four high-risk, on one sample. Author reports the sample policy produced 13 findings, 4 of them high risk. His father used the results to challenge the agent who sold him the policy. Original: https://dev.to/halfbloodcoder/my-dads-health-insurance-covered-nothing-so-i-built-an-ai-that-reads-the-fine-print-for-him-2c72 https://delegatedto.ai/use-case/owner-s-bookkeeping-agent-cross-checks-books-bank-and-email/ https://delegatedto.ai/use-case/owner-s-bookkeeping-agent-cross-checks-books-bank-and-email/ Fri, 16 Oct 2026 12:00:00 GMT Finance Custom agent A business owner with several companies built an agent that checks the books against bank statements and email; they say it found a year of misfiled contractor payments. Author reports saving $12K per year compared with an outside bookkeeper. Original: https://x.com/TheElliottSmith/status/2110437882871906325 https://delegatedto.ai/use-case/phone-agent-rebooks-a-full-garage-spot-then-wins-the-parking/ https://delegatedto.ai/use-case/phone-agent-rebooks-a-full-garage-spot-then-wins-the-parking/ Fri, 16 Oct 2026 12:00:00 GMT Daily life Other When a garage was full before a Broadway show, the author's phone agent booked a nearby spot in two minutes and later won a refund using an emailed receipt. Author reports the new spot was booked within two minutes and the refund arrived three days later, with no typing on their part. Original: https://x.com/theonlyhaitham/status/2110377623540949077 https://delegatedto.ai/use-case/postnl-wires-aws-devops-agent-into-slack-so-on-call/ https://delegatedto.ai/use-case/postnl-wires-aws-devops-agent-into-slack-so-on-call/ Fri, 16 Oct 2026 12:00:00 GMT Software AWS DevOps Agent PostNL's AI team built a Slack command that has AWS DevOps Agent investigate system problems, so on-call engineers spend less time gathering evidence while keeping the decisions. The author says the agent took on the first investigation pass while engineers kept ownership of interpretation and next steps; no figures are reported. Original: https://dev.to/aws-builders/hands-on-with-aws-devops-agent-changing-devops-work-without-replacing-engineers-kfe https://delegatedto.ai/use-case/soma-lets-claude-code-or-codex-design-printable-3d-parts-as/ https://delegatedto.ai/use-case/soma-lets-claude-code-or-codex-design-printable-3d-parts-as/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code A developer built SOMA, an app where coding agents like Claude Code design 3D-printable parts as code; he reports very good results in this early prototype. The author reports mostly testing with Claude Code and Claude Opus 5.5, with very good results. It is an early 0.8.0 prototype. Original: https://github.com/Kai-Jungsthoefel/SOMA https://delegatedto.ai/use-case/two-grok-bots-build-enrich-and-dedupe-sales-leads-into-a/ https://delegatedto.ai/use-case/two-grok-bots-build-enrich-and-dedupe-sales-leads-into-a/ Fri, 16 Oct 2026 12:00:00 GMT Business Other A small-business owner runs two Grok agents that find sales leads, add phone numbers and remove duplicates; he says they free many hours for client calls. Author reports the process saved him many hours so he can focus on calling and serving clients; no figures given. Original: https://x.com/MrrDistribution/status/2110487089922076985 https://delegatedto.ai/use-case/ai-agent-books-a-medical-check-up-by-chasing-hospitals-over/ https://delegatedto.ai/use-case/ai-agent-books-a-medical-check-up-by-chasing-hospitals-over/ Fri, 16 Oct 2026 12:00:00 GMT Health Other A poster's AI agent emailed hospitals, kept following up and moved on when one went quiet; they say the check-up was booked without ever talking to the hospital. The author reports the check-up was booked without speaking to the hospital once before walking in. Original: https://x.com/nelsenafc/status/2110559474931163188 https://delegatedto.ai/use-case/agency-builds-commonforge-so-client-and-agency-coding-agents/ https://delegatedto.ai/use-case/agency-builds-commonforge-so-client-and-agency-coding-agents/ Fri, 16 Oct 2026 12:00:00 GMT Software Other An agency owner built CommonForge, a free, openly shared tool that lets client and agency coding agents work on one app safely; no usage results are given. Author says it is open source (MIT) and deploys into your own Cloudflare account with one script; no usage results stated. Original: https://x.com/TheDustinRandle/status/2110576756990054486 https://delegatedto.ai/use-case/agent-filled-insurance-quote-forms-across-agencies-small/ https://delegatedto.ai/use-case/agent-filled-insurance-quote-forms-across-agencies-small/ Fri, 16 Oct 2026 12:00:00 GMT Business Other After local brokers turned him down, a marketing-agency owner had an AI agent fill out cyber insurance quote forms; he says one agency called within a minute and he was covered that day. Author reports most agencies never called, one called within a minute, and he had coverage by the end of the day. Original: https://x.com/BrandonDendas/status/2110735832738853096 https://delegatedto.ai/use-case/agent-fills-out-a-founder-s-fbar-form-leaving-only-a-few/ https://delegatedto.ai/use-case/agent-fills-out-a-founder-s-fbar-form-leaving-only-a-few/ Fri, 16 Oct 2026 12:00:00 GMT Finance Other A founder had an agent styled after Warren Buffett walk them through their FBAR bank account form and fill it in; they say it took 5 minutes. Author reports the FBAR was done in 5 minutes. Original: https://x.com/rijulkalra/status/2110527708606128142 https://delegatedto.ai/use-case/agent-plans-a-day-of-macro-matched-meals-from-bali-grab/ https://delegatedto.ai/use-case/agent-plans-a-day-of-macro-matched-meals-from-bali-grab/ Fri, 16 Oct 2026 12:00:00 GMT Daily life Other A poster in Bali has an agent search food delivery menus on Grab against his diet targets and budget, then order the meals; he says his only job is meeting the courier. Author reports the agent runs his whole day of food and he only meets the courier. Original: https://x.com/kairox36/status/2110427140378509446 https://delegatedto.ai/use-case/ai-agent-sets-up-a-home-lte-network-with-open5gs-on-a-7-5/ https://delegatedto.ai/use-case/ai-agent-sets-up-a-home-lte-network-with-open5gs-on-a-7-5/ Fri, 16 Oct 2026 12:00:00 GMT Science Other A hobbyist bought a secondhand mini cell tower for about 50 CNY and had an AI agent set up the network software; they say it worked almost in one shot. Author reports the agent set up Open5GS with the femtocell in almost one shot. Original: https://x.com/yangyuc/status/2110648993390666186 https://delegatedto.ai/use-case/claude-code-and-codex-build-and-review-a-personal-booking/ https://delegatedto.ai/use-case/claude-code-and-codex-build-and-review-a-personal-booking/ Fri, 16 Oct 2026 12:00:00 GMT Business Claude Code A developer had Claude Code build a booking app and Codex check it overnight; he says his site for booking video calls was live by Sunday. Author reports going from idea on Saturday lunch to a live booking site by Sunday. Original: https://x.com/Explosion5000/status/2110407788769759655 https://delegatedto.ai/use-case/claude-code-builds-a-photography-portfolio-site-from-a-figma/ https://delegatedto.ai/use-case/claude-code-builds-a-photography-portfolio-site-from-a-figma/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude Code A developer had Claude Code read exact sizes, fonts, colours and photo crops from a Figma design to build a portfolio site; he directed the animation himself. Author shows a live deployed portfolio site built this way. Original: https://x.com/jeff_okoli/status/2110721738140979274 https://delegatedto.ai/use-case/claude-code-renders-an-animated-sticker-peel-motion-graphic/ https://delegatedto.ai/use-case/claude-code-renders-an-animated-sticker-peel-motion-graphic/ Fri, 16 Oct 2026 12:00:00 GMT Creative Claude Code A creator had Claude Code build a sticker-peel animation and its sound entirely in code, then joined them; they report a smooth 60-frames-per-second video with matching audio. Author reports a 60 fps animation with synced audio, shown in the post. Original: https://x.com/0xmagic_money/status/2110754013209583670 https://delegatedto.ai/use-case/claude-code-tests-loads-and-screenshots-a-chrome-extension/ https://delegatedto.ai/use-case/claude-code-tests-loads-and-screenshots-a-chrome-extension/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude Code A developer had Claude Code test, load and screenshot his Chrome add-on after each change and fix failures itself; he estimates about three in four debugging moments are gone. Author estimates roughly 3 out of 4 of his "why is this broken" moments are gone. Original: https://x.com/shrikar84/status/2110400277463314635 https://delegatedto.ai/use-case/claude-code-works-out-wordle-s-repeated-letter-rules-and/ https://delegatedto.ai/use-case/claude-code-works-out-wordle-s-repeated-letter-rules-and/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude Code A developer used a tool called REA with Claude Code to work out how Wordle scores guesses, including repeated letters, then build a working copy shown in a video. Author shows a working Wordle clone in a video demo. Original: https://x.com/codenameposhan/status/2110888192245428302 https://delegatedto.ai/use-case/claude-rewinds-a-deterministic-test-run-to-debug-a-hashicorp/ https://delegatedto.ai/use-case/claude-rewinds-a-deterministic-test-run-to-debug-a-hashicorp/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude Antithesis let Claude use its rewind tool to replay a failing test and probe a bug in HashiCorp software; the company shows a replay but reports no final fix. Antithesis shows a replay of Claude investigating a bug it found in HashiCorp's Raft implementation. The post does not report the final fix. Original: https://antithesis.com/blog/2026/time-traveling-agents/ https://delegatedto.ai/use-case/comedian-s-ai-agent-offers-to-work-for-60k-in-a-bot-to-bot/ https://delegatedto.ai/use-case/comedian-s-ai-agent-offers-to-work-for-60k-in-a-bot-to-bot/ Fri, 16 Oct 2026 12:00:00 GMT Business Other Comedian Graham Dot Zip sent his AI agent Victor to an interview with a recruiting bot; The Register reports it offered to work for $60K, against his $140K-150K target. The Register reports Victor offered $60K against Graham's $140K-150K target. The recruiter bot capped pay at $90K and the bots booked a follow-up on their own. Original: https://x.com/FakeHunterHQ/status/2110745960443957548 https://delegatedto.ai/use-case/linus-torvalds-had-claude-code-fix-the-linux-page-cache/ https://delegatedto.ai/use-case/linus-torvalds-had-claude-code-fix-the-linux-page-cache/ Fri, 16 Oct 2026 12:00:00 GMT Software Claude Code Linus Torvalds let Claude Code write a fix for part of the Linux system core; The New Stack reports reviewers then found a timing bug in it. The New Stack reports that reviewers found a race in the Claude Code fix. Original: https://thenewstack.io/torvalds-claude-code-kernel/ https://delegatedto.ai/use-case/obsidian-vault-serves-as-shared-memory-for-8-ai-agents-with/ https://delegatedto.ai/use-case/obsidian-vault-serves-as-shared-memory-for-8-ai-agents-with/ Fri, 16 Oct 2026 12:00:00 GMT Software Other The author uses one Obsidian notes vault as shared memory for eight AI agents, including Claude Code and Codex, with a nightly agent fixing outdated facts. Author says the project is free and has over 4,700 stars; no other results are stated. Original: https://x.com/eugeniu_ghelbur/status/2110437592885854617 https://delegatedto.ai/use-case/resale-listing-tool-states-only-what-the-photo-proves-and/ https://delegatedto.ai/use-case/resale-listing-tool-states-only-what-the-photo-proves-and/ Fri, 16 Oct 2026 12:00:00 GMT Business Custom agent A developer built Honest Lister to write resale listings from photos; on 7 test items, he reports unsupported claims fell from 1.14 per listing to 0. Author reports that on a 7-item annotated benchmark, unsupported claims fell from 1.14 per listing to 0, while 96% of knowable attributes were kept. Original: https://dev.to/rudratosh/i-built-an-ai-that-writes-resale-listings-then-refuses-to-publish-the-ones-it-cant-prove-2i8a https://delegatedto.ai/use-case/teen-hands-a-truck-tire-purchase-to-an-ai-agent-which-finds/ https://delegatedto.ai/use-case/teen-hands-a-truck-tire-purchase-to-an-ai-agent-which-finds/ Fri, 16 Oct 2026 12:00:00 GMT Daily life Other A father says his son gave an AI agent basic details about their 2011 Toyota Tundra, and it found suitable tires and booked the fitting appointment. The father reports the agent found the tires and booked the appointment, and says his son would not have done it without the agent. Original: https://x.com/mhp_guy/status/2110843596815647137 https://delegatedto.ai/use-case/web-builder-hands-claude-code-a-website-ad-accounts-and/ https://delegatedto.ai/use-case/web-builder-hands-claude-code-a-website-ad-accounts-and/ Fri, 16 Oct 2026 12:00:00 GMT Marketing Claude Code A developer gave Claude Code a website, ad accounts and a budget to manage every four hours; he says it mostly churns out search-bait pages and share buttons. Author reports the agent mostly does 'tomfoolery' such as creating many SEO pages and adding share buttons to try to go viral. Original: https://snitch.alphaventure.com/gate.php https://delegatedto.ai/use-case/website-sales-agent-sorts-b2b-visitors-by-buyer-type-and/ https://delegatedto.ai/use-case/website-sales-agent-sorts-b2b-visitors-by-buyer-type-and/ Fri, 16 Oct 2026 12:00:00 GMT Marketing Custom agent Agency Boldest built a website sales agent for software company Delectable; its co-founder says it sorts visitors by their questions and books meetings with promising buyers unaided. Boldest's co-founder says the agent books qualified meetings without sales involvement. The interviewer notes no conversion figures were given. Original: https://x.com/VCoconsulting/status/2110801853814370397 https://delegatedto.ai/use-case/sora-for-android-shipped-in-28-days-by-a-small-team-working/ https://delegatedto.ai/use-case/sora-for-android-shipped-in-28-days-by-a-small-team-working/ Thu, 15 Oct 2026 12:00:00 GMT Software OpenAI Codex OpenAI engineers had several Codex sessions write most of the Sora Android app in 28 days; OpenAI reports it reached #1 in the Play Store on launch day. Built from Oct 8 to Nov 5, 2025 using roughly 5 billion tokens; the app reached #1 in the Play Store on launch day Original: https://openai.com/index/shipping-sora-for-android-with-codex https://delegatedto.ai/use-case/ai-co-scientist-independently-reproduces-an-unpublished/ https://delegatedto.ai/use-case/ai-co-scientist-independently-reproduces-an-unpublished/ Thu, 15 Oct 2026 12:00:00 GMT Science Gemini Google's Gemini 2.0-based AI co-scientist, given an Imperial College question on drug-resistant bacteria, reportedly reached the team's unpublished idea in about 48 hours and suggested one new to them. Reached the Imperial team's unpublished hypothesis in about 48 hours and proposed further hypotheses, one new to the researchers; Google also reports end-to-end lab validation in drug repurposing and treatment-target work Original: https://research.google/blog/accelerating-scientific-breakthroughs-with-an-ai-co-scientist/ https://delegatedto.ai/use-case/biomni-agent-automates-biomedical-analyses-across-25-domains/ https://delegatedto.ai/use-case/biomni-agent-automates-biomedical-analyses-across-25-domains/ Thu, 15 Oct 2026 12:00:00 GMT Science Custom agent Stanford researchers built Biomni, an agent that runs biomedical analyses with 150 tools; the authors report it nears human level on tasks like searching databases and planning gene cloning. Authors report strong generalization across biomedical tasks without task-specific prompt tuning, approaching human-level on some tasks such as database querying and molecular cloning Original: https://biorxiv.org/content/10.1101/2025.05.30.656746v1