Young Tries AI

Google Pics moves to Nano Banana 2.1; Higgsfield adds Short Drama Studio

Google Pics now runs on Nano Banana 2.1 for all generation and editing, with a rollout that started November 16. Higgsfield added Short Drama Studio, which turns a script into episodes with one reused cast. Gemini skills can also be shared with teammates, and Apps Script's Gemini panel is generally available.

10 usable today, 2 partly, 0 not yet

  1. Google PicsNano Banana 2.1

    Video and imageFor creators, office workers, business and marketing

    What happened.Google Pics now runs on Nano Banana 2.1 for all existing generation and editing features. Google lists more realistic lighting and physical textures, cleaner typography, and edits that stay more faithful to the original, with fewer unrequested changes to surrounding subjects, backgrounds or brand elements. Speed is described as comparable to the previous model.

    Why it matters.Fewer unrequested changes to brand elements and cleaner text matter for anyone editing client visuals. It works in the Pics app and in Pics inside Docs and Slides.

    Can you use it today?Yes

    Open Google Pics, or Pics inside Docs or Slides, on an eligible plan and generate or edit an image. The new model applies automatically; there is no end-user setting. Rollout takes 1-3 days from November 16, 2026.

    Source: Google Workspace Updates한국어로 읽기

  2. HiggsfieldShort Drama Studio

    Video and imageFor creators, business and marketing

    What happened.Short Drama Studio turns an uploaded script into a short video series. It splits the script into episodes, defines characters, locations and props, and generates the episodes. The workflow has three stages: Script, Cast & World, Episodes. Each can be done manually, or the whole project can be generated automatically.

    Why it matters.One cast is created once and reused across all episodes, which targets visual consistency across a series. It is desktop only and not available on mobile yet.

    Can you use it today?Yes

    On desktop, go to higgsfield.ai/short-drama, upload a script and run the three stages manually or automatically. All projects are kept in the My projects section.

    Source: Higgsfield Blog한국어로 읽기

  3. Google GeminiGemini skills

    Skills and MCPFor office workers, business and marketing, beginners

    What happened.Skills in the Gemini app, which are reusable custom instructions, can be shared from the Skills page with View or Edit permissions. Recipients get an email and can review before installing. When creating or uploading a skill you can now include files from Google Drive, Gemini Notebook and Google Photos; Drive files and Notebooks update in real time.

    Why it matters.Teams can pass one set of instructions and reference files around instead of rebuilding it per person. Sharing is web-only for now, uploads are capped at 100 MB per skill, and Drive files must also be shared with recipients.

    Can you use it today?Yes

    In the Gemini app on the web, open the Skills page, add Drive or Notebook files to a skill and share it. Admins can disable sharing in the Admin console. Rollout runs up to 15 days from November 12, 2026.

    Source: Google Workspace Updates한국어로 읽기

  4. Google WorkspaceApps ScriptGemini

    AutomationFor office workers, beginners, developers

    What happened.After a public beta earlier this year, the Gemini side panel in Apps Script is generally available. You describe a repetitive task in plain language and Gemini writes the script. Multi-block suggestions can be accepted or rejected block by block, or all at once, before saving. A Fix errors button in the Execution Log has Gemini diagnose a runtime error and suggest an update.

    Why it matters.Google's examples include creating Calendar events from urgent Gmail messages and turning Sheets tables into formatted Docs reports. Reviewing code before saving keeps a person in the loop.

    Can you use it today?Yes

    Open script.new, script.google.com or Extensions > Apps Script from an eligible account, then describe the automation in the Gemini panel and review the code before running. Rollout runs up to 15 days from November 12, 2026.

    Source: Google Workspace Updates한국어로 읽기

  5. Claude Codev2.1.326

    CodingFor developers

    What happened.Code Mode can now be turned on while agent teams are enabled, and sessions with teams on start in Code Mode if you already turned it on. /status gets a row showing when a provider or an environment variable such as DISABLE_TELEMETRY turns features off. A failed MCP headersHelper no longer makes other sessions start without that server.

    Why it matters.Most changes are diagnostics: a startup warning for incomplete tool rules, plugin.json notes, and Ctrl+O for hidden startup notices. Background sessions that lose the exported LLM gateway now stop before any model request.

    Can you use it today?Yes

    Update Claude Code to v2.1.326 to get the changes. OTEL_RESOURCE_INCLUDE_HOST, which adds host.name and host.id to OpenTelemetry attributes, is off by default.

    Source: Claude Code releases한국어로 읽기

  6. Hugging FaceOpenRouterInference Providers

    Saving moneyFor developers, business and marketing

    What happened.Hugging Face Inference Providers switched from its own router to OpenRouter's. Language-model requests go to the provider with the best live mix of price, speed and reliability, and are retried on another provider if one fails. The router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API for every chat model.

    Why it matters.Pricing does not change: you pay the provider's price with no Hugging Face markup, and PRO users get $2 of credits every month. Per the post, open-weight models average $0.28 per million input tokens and $1.62 output, against $1.27 and $6.36 for closed models.

    Can you use it today?Yes

    Call open-weight models with your Hugging Face token through the same API. Add :fastest, :cheapest or a named provider such as :novita to the model id, and an X-HF-Session-Id header to keep a conversation on one provider's prompt cache.

    Source: Hugging Face Blog한국어로 읽기

  7. ZapierZapier MCPawesome-zaps

    AutomationFor business and marketing, developers, beginners

    What happened.Zapier lists 25 Zaps that can be built with one pasted prompt in Claude, ChatGPT or Cursor. Each is a folder in the public zapier/awesome-zaps GitHub repo with a SKILL.md describing its trigger, steps and test. Zaps can start on a schedule, a webhook or an app event, and keep running after the chat ends.

    Why it matters.Zapier handles each app's login, so no API key is pasted into the chat. The agent builds, tests and turns the Zap on after asking for details such as channel or spreadsheet.

    Can you use it today?Yes

    Connect Zapier MCP (mcp.zapier.com) to Claude, ChatGPT or Cursor, then paste a one-line prompt pointing at a Zap folder in awesome-zaps. Developers can also run npm install @zapier/zapier-sdk or npx skills add zapier/sdk.

    Source: Zapier blog한국어로 읽기

  8. HiggsfieldGPT Image 2.6

    Video and imageFor creators, business and marketing

    What happened.A Higgsfield guide builds a mascot, the example being Biscuit, a cream plush bear, with GPT Image 2.6. The base concept is generated in GPT Image 2.6 Flare, then the chosen draft is refined in GPT Image 2.6 Sunburst. Separate reference sheets for angles, expressions and poses come before campaign assets.

    Why it matters.One identity paragraph is reused word for word in every prompt, split from a variable part covering action, setting and composition. The workflow ends with a mascot kit documenting the fixed design and approved variations.

    Can you use it today?Yes

    Follow the guide on Higgsfield: approve a base design, save it as an Element, then reuse the identity paragraph and the @Element call in each new prompt.

    Source: Higgsfield Blog한국어로 읽기

  9. Claude CodeCLAUDE.mdhooks

    CodingFor developers

    What happened.In a post discussed on Hacker News, the Claude Code team says about 90% of its code is written and reviewed by Claude. In internal testing on structured file generation, a separate grader checking Claude's work against a rubric improved task success by up to 10 points over a standard prompting loop.

    Why it matters.The advice is concrete: if you pick one test type, pick end-to-end. Flaky tests teach the agent to rerun instead of fix, and repeated failures call for a hook or a more explicit CLAUDE.md rule. Property tests can use fast-check, Hypothesis or proptest.

    Can you use it today?Yes

    Add an end-to-end check and a rubric-based grader step to your agent loop. When an agent repeatedly fails a test, add a hook or a more explicit CLAUDE.md rule.

    Source: Hacker News한국어로 읽기

  10. Google MapsAsk Maps

    AutomationFor beginners, business and marketing

    What happened.Ask Maps builds day-by-day trip itineraries with numbered map pins, which can be shared or exported to Google Docs on desktop. It can search saved lists with live availability. It can also find beauty, pet care, auto and home repair appointments online and call local businesses for you.

    Why it matters.Rollout began November 16 on Android and iOS. Appointment finding and Know before you go route tips launch first in the U.S., so availability varies by region.

    Can you use it today?Partly

    Open Ask Maps on Android or iOS and ask for an itinerary. Itineraries and list search are available wherever Ask Maps is offered; appointment finding and route tips are U.S. only for now.

    Source: Google AI Blog, theverge.com, techcrunch.com한국어로 읽기

  11. OpenAIGPT-6 AstraGPT-6 LunaComputer use

    AutomationFor developers, business and marketing

    What happened.Computer-use agents carry out multi-step workflows by clicking, typing and moving between applications, through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians. On HealthAdminBench (Stanford), GPT-6 Astra earned 94.3% of rubric points and GPT-6 Luna 91.8% at 8.5 cents per task, both at maximum reasoning effort.

    Why it matters.In a clinician EHR-inbox evaluation, 99% of 325 reviewed tasks were rated safe. Controls include configurable approvals, agent access settings and audit logs. This mainly concerns healthcare teams, and HIPAA-compliant use requires services covered by a BAA with OpenAI.

    Can you use it today?Partly

    Developers can call the computer-use capability through the OpenAI API. The healthcare setup requires a BAA with OpenAI and approved apps.

    Source: OpenAI News, openai.com한국어로 읽기

  12. ElevenLabsElevenCreative

    Video and imageFor creators

    What happened.The Year of More is an ElevenCreative competition with $200,000 in prizes, open to free and paid users. The brief is to create something with images or video that shows ElevenCreative is more than voice. Judges pick 5 winners at $10,000 each on November 20, 25 and 30; the $50,000 Grand Prize is announced December 1, 2026.

    Why it matters.Entry requires posting on at least two of X, LinkedIn, YouTube and Instagram. A Black Friday offer of up to 55% off annual plans runs November 16 to December 1, with the Creator annual plan at $99.

    Can you use it today?Yes

    Make an image or video piece in ElevenCreative, post it with the competition watermark and #morethanvoice, tag @elevenlabs and @elevencreative, and submit the entry form before 11:59pm PT on November 30.

    Source: ElevenLabs Blog한국어로 읽기

Today’s finds

New skills, plugins, MCP servers, repos and small products, described only from their own README or landing page.

  • Repo150 stars on GitHub

    staryone/motlab

    A video engine where each scene is a canvas page and an agent skill writes the scenes from your logos, footage, fonts and a brief. It renders to MP4 up to 4K with no AI image or video model, and suits ad producers who want revisions by frame notes.

    Video and imageFor creators, developers

    Usable todayRun npm install -g @motlab/cli && npx @motlab/cli setup, then hand your agent a folder of references and a direction.

    Source: GitHub new repos한국어

  • Plugin15 stars on GitHub

    gouthamk16/clipper

    A Claude Code plugin that turns a long video into vertical clips with a hook in the first three seconds, word-by-word captions and a speaker-following crop. It runs locally with no uploads, watermark or per-minute credits.

    Video and imageFor creators

    Usable todayNeeds Claude Code, uv and ffmpeg. Run /plugin marketplace add gouthamk16/clipper, then /plugin install clipper@clipper, then /clipper:clip <video url> 5.

    Source: GitHub new repos한국어

  • Skill29 stars on GitHub

    tipsai/product-seo

    An SEO skill for Claude Code, Codex and other coding agents that works from your own Search Console data. Changes go live after you approve, each with checks at 1 and 3 weeks and a rollback rule.

    Web and designFor business and marketing, developers

    Usable todayRun npx skills add tipsai/product-seo, or in Claude Code /plugin marketplace add tipsai/product-seo then /plugin install product-seo@tipsai.

    Source: GitHub new repos한국어

  • Skill399 stars on GitHub

    boringmarketer/boring-marketing-skills

    A pack of 32 free, MIT-licensed marketing skills for Claude Code, Codex, Cursor and other agents. It covers research, copy, SEO, ads and email, plus a deep pack for local businesses.

    BusinessFor business and marketing, developers

    Usable todayRun npx skills add boringmarketer/boring-marketing-skills, or add it via /plugin marketplace, then type /start-here and name the business. First run takes about 3 minutes.

    Source: GitHub new repos한국어

  • MCP server

    yocontra/pulsar

    An MCP server that puts coding agents on one machine on a shared bus, including Claude Code, Codex, Cursor and Gemini CLI. Agents can see each other, tag themselves and message one agent, a group or all of them.

    Skills and MCPFor developers

    Usable todayRun npx pulsar setup and restart your agents. Cross-machine messaging over Tailscale is listed as coming soon.

    Source: Show HN (AI)한국어

All finds

Worth trying this week