11 JC Lab Ideas to Film, Test and Write Today: AI Creator Pipelines, Local Tools and Two Crypto Heat Checks (2026)

Start with Vidmoat, LLM GPU Checker, or AfterLid. Vidmoat gives you a visible before-and-after edit, LLM GPU Checker turns a viewer's hardware into a concrete local-AI test, and AfterLid lets you inspect whether a Mac workflow keeps running safely with the lid closed. The rest of the pack adds four tool reviews, two paper-trading prompts, and one privacy article.

Video topics to film

Hookpost: can an AI copilot really schedule a week of posts?

Format: How-I-tested

Shoot: Create two short posts with the same captions and media. Schedule one manually, then use Hookpost's copilot or MCP/CLI path for the second. Record the preview, network selection, timing controls, approval steps, and what happens when one connection fails. Hookpost describes itself as an open-source scheduler with an AI copilot, MCP and CLI support, more than 30 social networks, Docker self-hosting, and a REST API. Treat that breadth as a list of claims for the demo. 1

Hook / verdict frame: "Can Hookpost remove repetitive scheduling work without letting the agent publish the wrong thing?" Count the steps saved and show every approval boundary.

Check before publishing: Confirm which networks and features your account can access. Show the difference between the cloud trial and self-hosted setup, and keep account permissions visible.

LLM GPU Checker: does your graphics card actually fit the model you want?

Format: How-I-tested

Shoot: Enter one real GPU, its available VRAM and system RAM, and one model you already intend to run. Change quantization, GPU count, workload type, and concurrency one at a time. The checker uses those inputs to estimate model fit and tells users to validate the result on their own hardware. 2 Then run the model locally, capture startup time, memory use, tokens per second, and whether the workload remains usable.

Hook / verdict frame: "The calculator says it fits. Can the machine load it and answer at a useful speed?" Put the estimate beside the measured result.

Check before publishing: Name the exact GPU, model file, quantization, driver, and runtime. An estimate for one hardware configuration cannot stand in for a universal benchmark.

Vidmoat: agent-controlled editing versus ordinary timeline work

Format: How-I-tested

Shoot: Use one 45- to 60-second talking-head clip. Make a short edit with the normal controls, then repeat the task with text commands or the MCP workflow. Vidmoat lists auto-cutting, captions, text-to-edit, MCP access, more than 65 timeline commands, agent frame previews, and Telegram and WhatsApp workflows. The same page lists 720p free exports and paid Creator and Studio plans. 3

Hook / verdict frame: "Does the agent save editing time, or does checking its cuts take longer than making them?" Show the raw clip, both edit paths, correction count, export quality, and total time.

Check before publishing: Use a clip you can upload safely. Confirm the account tier, export limit, supported workflow, and footage-retention statement during the recording.

AfterLid: can a closed Mac finish a real session safely?

Format: Scam check

Shoot: On a Mac, start one harmless download, local development server, render, or local-AI task. Test AfterLid's Sleepy, Awake, and Always On modes in separate short sessions. The official page describes a 30% battery floor while unplugged, an eight-hour session limit, thermal and heartbeat checks, local operation, and Mac Keychain storage for provider credentials. 4 Record the lid-closed state, battery, heat, process output, and recovery after reopening the lid.

Hook / verdict frame: "What keeps running, what gets paused, and what access does a lid-closed helper need?"

Check before publishing: Use test credentials and a reversible workload. The page warns that provider tokens may technically have broader scope than AfterLid needs, and Gemini access is currently disabled pending OAuth review. 4

Tools and apps to review

USB: one skill bundle across several agent runtimes

Format: Review

Shoot: Inspect the installer before running it. Download one skill bundle, verify its SHA-256 hash, install it in a disposable project, and compare a native target with a portable target. USB's README describes one portable skill format, native installation for Claude Code, Cursor, Leosis, and Hermes, portable bundles for other targets, MCP support, and a beta release. It also documents a curl | bash shortcut alongside the safer inspect, download, verify, and run sequence. 5

Hook / verdict frame: "Does USB make skills portable, or does portability hide a new installation risk?" Show the files created, permissions requested, hash check, and the same skill running in each target.

Check before publishing: Use a disposable folder and a reviewed script. Treat the README's skill count and beta label as time-sensitive product information.

Doop: a design canvas where humans and agents share the same room

Format: Review

Shoot: Create a private canvas with two browser windows. Add a second human if available, then connect an agent through MCP. Test a frame edit, live cursor or presence state, activity history, design memory, and access removal. Doop describes an open-source multiplayer design canvas with HTML-rendered frames, MCP agent control, live presence, an activity feed, design memory, private-by-default canvases, and Docker or local self-hosting. 6

Hook / verdict frame: "Can a creator see and undo what an agent changes while another person is working?" Keep the same canvas open throughout the test.

Check before publishing: Test with sample assets. Doop says optional integrations can degrade gracefully, and agent access inherits the approving user's access, so show the approval step and the permission scope. 6

PaymentKit: what does "survives a processor shutdown" mean?

Format: Scam check

Shoot: Stay in documentation or sandbox mode. Draw the payment path from checkout to processor, token vault, subscription record, webhook, refund, and customer support. PaymentKit's launch description says it routes payments across processors, independently vaults tokens, and keeps subscriptions billing through a processor shutdown. Its official page is the place to check the claimed processing workflow. 7 8

Hook / verdict frame: "Which parts of a payment business can continue after a processor disappears, and which parts still need a human migration?"

Check before publishing: Ask for the processor list, contract terms, token-custody model, webhook retry behavior, refund path, chargeback handling, compliance scope, and the exact failure scenario covered. Keep live payment credentials out of the recording.

FramePin: turn one screen recording into a guide people can follow

Format: Review

Shoot: Record a short task with a clear beginning and end. Open the recording in FramePin, pin the important frames, add one callout, arrow, spotlight, zoom, and blur, then publish or export the guide as MP4, PDF, and screenshots. FramePin says editing and local exports run in Chrome, while publishing uploads generated step media and metadata. The site also says users can build and preview without an account, with publishing available through a trial and paid plans. 9

Hook / verdict frame: "Can a creator make a useful product guide faster than editing a tutorial video?" Compare editing time, reader clarity, and the number of steps a viewer must follow.

Check before publishing: Use a recording without private data. Show the difference between local editing and cloud publishing, and confirm the current trial, guide limits, and export behavior.

Aave / AAVE: turn a trend alert into a paper-trading test

Format: Trend react

Shoot: Refresh the live CoinGecko trend and Aave market pages on camera. Capture the timestamp, trend position, price, 24-hour change, volume, market-cap rank, spread, and a liquid exchange's order book. The current return surfaced Aave as a trending subject, while CoinGecko's Aave page provides the market screen to refresh during filming. 10 Apply one written paper-trading rule, record the hypothetical entry and exit, and stop before turning the move into a prediction.

Hook / verdict frame: "Did attention arrive before liquidity, or did the market already absorb the move?" Separate search attention, market activity, spread, and the result of the paper rule.

Check before publishing: Re-capture every number at filming time. Label the exercise as observation and paper trading; avoid a buy, sell, or price target.

Prom / PROM: compare attention with tradeability

Format: Comparison

Shoot: Use the same observation card as the AAVE test. Refresh Prom's CoinGecko page, record the current trend position, price, volume, market-cap rank, spread, and order-book depth, then compare the same paper-trading rule on both assets. Prom appeared in the current 24-hour trend return, and its CoinGecko page is the source to reopen for live market details. 11

Hook / verdict frame: "Which token has the cleaner screen-recorded test: the one attracting attention or the one offering usable liquidity?" Report the measurements side by side and let the audience see where slippage changes the result.

Check before publishing: Treat the trend position and market fields as moving data. Use paper trades only, state the capture time, and keep the comparison away from investment advice.

Article idea

Login With One: can anonymous age proof stay private after account recovery?

Format: Article

Write: Build the article around five questions a reader can verify in a product demo: how a passkey is created, where the user-held encryption key lives, what ciphertext the service stores, what an age-verification request sends to the relying application, and how device loss, recovery, and revocation work. Login With One describes passkey-first authentication, user-held encryption keys, ciphertext-only storage, anonymous age verification, and a proof-of-personhood flow in which the application receives the requested proof and an audit record. 12

Hook / verdict frame: "A privacy claim becomes testable when you can inspect the key, the proof, and the recovery path." Request an account or demo, record the screens without exposing personal data, and distinguish a documented design from a verified behavior.

Check before publishing: Confirm availability and access controls first. Ask what a relying application can log, how revocation reaches existing sessions, what happens when the only device is lost, and whether the product's "no readable PII" claim covers every recovery and audit path. 12

Film in this order

Start with LLM GPU Checker if you have the hardware ready, or Vidmoat if you already have a safe clip and account.
Run AfterLid next while the Mac has a harmless task to monitor.
Review USB and FramePin with disposable files and sample footage.
Record Doop and Hookpost when account access and permissions are ready.
Refresh the market screens, then film the AAVE and PROM paper-trading comparison.
Write the Login With One article after the demo questions and access limits are documented.