CASE 01DemoTrain a Personal Model From iMessage
Use GPT-5.6 to build and run a local training pipeline that learns a personal writing style from private message history.
@skiranoRECENTLY UPDATED / (SHANGHAI TIME)
CHEER LIN / ARCHIVE
Browse real GPT-5.6 work across autonomous coding, agent collaboration, creative builds, and capability limits.
USE CASE MAP / 01
This collection preserves the key result, evidence type, creator, and original source for 167 public cases so you can quickly spot repeatable methods.
HOW TO USE / 02
Narrow by task, verify context at the original source, then reproduce the most useful path through EvoLink.
Start with coding, agents, creative work, evaluation, or another task family.
Use titles, methods, or creators to find the closest case to your current goal.
Select the creator to review the full demo, limits, context, and original explanation.
Choose a method worth testing, connect to the model, and turn it into your workflow.
REAL CREATOR CASES / 03
Up to five cases per row, with fewer columns on narrower screens. Select a creator to open the original case; images and videos load only when needed.
CASE 01DemoUse GPT-5.6 to build and run a local training pipeline that learns a personal writing style from private message history.
@skirano
CASE 02DemoGive a long-running coding agent a voxel Manhattan build and let it work autonomously over multiple days.
@mattshumer_
CASE 03DemoTurn a spoken natural-language specification into an end-to-end service build.
@OpenAIDevs
CASE 04TutorialStart with one repetitive task, grant Codex access to the relevant tools, and expand the workflow after it works reliably.
@gregisenberg
CASE 05DemoUse a second agent as a critique pass when reviewing GPT-5.6's own work.
@AdamHoltererer
CASE 06TutorialDrop an MP4 into the workspace and request a concise promotional cut in natural language.
@clairevo
CASE 07IntegrationConnect GPT-5.6 to an ad-production MCP so the agent can use brand context, video tools, and reusable skills in one workflow.
@Creatify_AI
CASE 08IntegrationUse GPT-5.6 in Figma Make when building from an existing design, then compare output strength and token efficiency.
@figma
CASE 09BenchmarkUse third-party benchmark results to choose Sol, Terra, or Luna by intelligence, coding performance, and cost per task.
@ArtificialAnlys
CASE 10EvaluationBenchmark visual polish and physical correctness separately before choosing Sol Ultra for browser-based simulations.
@atomic_chat_hq
CASE 11DemoUse ChatGPT Work to act across apps and files, stay with a project for hours, and return finished work.
@OpenAI
CASE 12DemoDescribe an operational problem on site, then use GPT-5.6 to build a tracking and automation workflow around it.
@OpenAI
CASE 13BenchmarkCompare benchmark score, task cost, and token use together instead of ranking coding models on score alone.
@bridgemindai
CASE 14BenchmarkUse ARC-AGI-3 to test how well a model orients itself in unfamiliar interactive tasks.
@arcprize
CASE 15IntegrationUse GPT-5.6 inside Hermes Agent with Nous Portal as the model access layer.
@NousResearch
CASE 16DemoGive GPT-5.6 Sol a complete research prompt and use it to design the post-training process for GPT-5.6 Luna.
@sharifshameem
CASE 17BenchmarkCompare Sol, Terra, and Luna on the same software-engineering leaderboard before selecting a tier.
@datacurve
CASE 18EvaluationReview aggregate agent activity and completed projects before drawing conclusions from a large model trial.
@theo
CASE 19EvaluationUse one agent for small edits, browser operations, and multi-day builds, while retaining approval for consequential actions.
@MatthewBerman
CASE 20IntegrationUse the unified desktop app for inline edits, pull-request review, multi-repository work, and faster Computer Use.
@CodexReleases
CASE 21BenchmarkCompare agent benchmark score and estimated cost at the same reasoning setting.
@OpenAI
CASE 22DemoUse a long-running build to combine 3D terrain, satellite imagery, weather, search, and cinematic navigation.
@pankajkumar_dev
CASE 23BenchmarkEvaluate coding score together with output tokens, elapsed time, and task cost.
@OpenAI
CASE 24EvaluationUse GPT-5.6 for broad knowledge-work loops and implementation, while reserving other models for the hardest architecture work.
@danshipper
CASE 25IntegrationRoute build tasks between GPT-5.6 Sol and Fable 5 while keeping deployment infrastructure in one agent environment.
@bindureddy
CASE 26IntegrationMatch Sol, Terra, or Luna to long-running reasoning, everyday coding, or fast low-cost tasks in GitHub Copilot.
@github
CASE 27DemoHave the agent inspect its rendered output, catch visual or functional defects, and revise before handoff.
@OpenAI
CASE 28LimitTreat cybersecurity safeguards as a deployment constraint because long-form agentic attacks remained reachable in testing.
@alxndrdavies
CASE 29DemoTurn data and concepts into charts, walkthroughs, 3D models, simulations, mini-apps, and shareable Sites.
@ajambrosino
CASE 30DemoUse GPT-5.6 to create an initial editable presentation, then review structure and visual quality before delivery.
@super_bonochin
CASE 31IntegrationUse GPT-5.6 with hosted agents in Microsoft Foundry for managed enterprise agent workflows.
@Azure
CASE 32IntegrationApply one preferred model across Word, Excel, PowerPoint, Chat, and Copilot Cowork knowledge-work tasks.
@Microsoft365
CASE 33DemoTurn a startup idea into a working subscription product page, then inspect pricing and checkout states before launch.
@Akasheth_
CASE 34DemoDictate a product idea, then have the agent implement a canvas tool that converts arranged boxes into structured image prompts.
@btibor91
CASE 35DemoUse GPT-5.6 to structure long image-generation requirements before sending them to an unchanged image engine.
@fuuro_ito
CASE 36EvaluationInspect benchmark submissions for grader-specific shortcuts before accepting performance scores.
@elliotarledge
CASE 37EvaluationCompare low-angle lighting outputs side by side when evaluating visual detail.
@AIna_artmusic
CASE 38DemoGenerate slides, sheets, and documents as editable artifacts instead of flattened images.
@neelajj
CASE 39IntegrationSelect GPT-5.6 through AI SDK and expose the maximum reasoning setting in application code.
@aisdk
CASE 40DemoOrganize recommendations by neighborhood, mood, and price, then publish the result as a Site.
@danizeres
CASE 41DemoTurn a vague executive request into a working browser-based dashboard inside a JetBrains IDE.
@jetbrains
CASE 42DemoLet a coding agent continue expanding a playable game for several hours while monitoring its progress.
@WesRoth
CASE 43BenchmarkBenchmark both final score and elapsed time with and without subagents.
@nickbaumann_
CASE 44IntegrationUse ChatGPT Work to automate internal workflows, surface insights, and reduce operating cost.
@nvidia
CASE 45IntegrationRun GPT-5.6 inside Devin Desktop as part of an agentic software-development workflow.
@devindesktop
CASE 46LimitTreat published usage ranges as estimates because model choice, context, reasoning, and tool use change consumption.
@super_bonochin
CASE 47LimitRead a complete generated story before judging prose quality and whether the writing still reveals its AI origin.
@fkadev
CASE 48IntegrationUse Microsoft 365 context through WorkIQ while generating and editing knowledge-work outputs in Copilot.
@rrr_kgknkShowing 48 / 167
BUILD WITH EVOLINK / 04
Connect to the model through EvoLink and build from directions already demonstrated in public.