AI Daily: Runway in OpenAI’s dots and the overnight ad

An empty director’s chair in a dark film studio at night, facing a wall of small monitors glowing cyan, with faint blue dawn light from a high window.

Written by Mohammad Nasravi inنوشته شده توسط محمد نصراوی در

Runway now runs inside OpenAI’s dots, the always-on agents OpenAI launched on 29 September. Runway’s pitch: brief your dot before bed, and the finished spot is waiting in the morning. The tool can make the shots overnight. Deciding what the shots are is still the director’s job, and that part does not belong in a night shift.

Runway inside OpenAI’s dots: the overnight spot

OpenAI introduced dots on 29 September: agents powered by GPT-6 Astra, each with its own cloud computer, that keep working between conversations. They are rolling out to Pro and Business Premium users in supported markets. Pro excludes the EEA, Switzerland and the UK at launch (OpenAI release notes, 29 Sep). On 1 October Runway added itself to dots. Its changelog says the dot “plans the shots, Runway shoots them” (Runway changelog, 1 Oct). Runway’s own setup page says that once you add Runway to ChatGPT, your dot can use it with no extra setup (Runway for ChatGPT).

My read: “Plans the shots” is the line to read twice. On Sky Red Grape the character and all four locations were fixed in design sheets before a single shot was made. The start and end keyframes were locked for the two hardest moves. That is direction, and none of it would go to an overnight run. What can go overnight is coverage: variations of shots you have already designed. OpenAI says Custom Rules can make an action wait for approval. Put paid generation behind that rule before the first night, or you wake up to a credit bill and twenty versions of a shot you never asked for.

Seedance 2.5 Draft mode: rough first, finish later

On 2 October Runway’s Agent started generating with Seedance 2.5 in Draft mode on all plans. It makes fast 480p drafts, then enhances them after generation. Runway says Draft mode is suggested automatically and can be switched off, and that it makes iteration “faster and cheaper”. The changelog gives no price (Runway changelog, 2 Oct).

My read: This is the right order: an animatic before the finished film, like any production. Just know what 480p can and cannot tell you. On Sky Red Grape nine of the 27 seconds belong to the product. At 480p the label is a smudge. Judge the timing, camera and motion in the draft, and check the pack only in the final. Enhancement sharpens what is already in the frame. It will not fix a label the draft got wrong.

ChatGPT Try on: your product image becomes a reference

On 1 October ChatGPT added a “Try on” button to product listings for clothes and accessories. You take or upload a selfie, and ChatGPT Images renders the item on you. The reference photo is saved for later try-ons, and you can delete it in Settings. You can also save products to Favorites. It is live on mobile and web (OpenAI release notes, 1 Oct).

My read: Look at who controls the final frame. It is not the brand’s campaign shot any more. It is a render built from your listing image and a stranger’s selfie. The note says nothing about how a brand checks those renders or reports a bad one. This first release covers clothing only, but FMCG should still take note: the product photo in your feed is now raw material. A clean, front-on, true-colour pack shot is worth more than another styled hero.

Pope Leo XIV: “Algorithms lack the spark of humanity”

On 2 October Pope Leo XIV wrote on X that it is “becoming urgent” to tell human art apart from machine output. He called the difference “ontological” and said “algorithms lack the spark of humanity”. He also said the Church wants a renewed alliance with artists and cultural institutions. The post had more than 24 million views by Monday (@Pontifex, 2 Oct).

My read: Read it as a sign, not a policy. For a large audience “made by a machine” now carries a moral meaning, not just a technical one. The risk is highest on sacred and seasonal occasions, because there the audience is judging sincerity. I made the Shirin Asal Christmas greeting end to end with Nano Banana, Kling and CapCut. I wrote the story and directed every beat myself. For a greeting, that human authorship is the brief, not a credit line. Decide before production what you disclose and where the human hand shows.

What I’d do today

Before any agent touches a paid generator, write the shot list yourself and lock the keyframes for the hardest moves. Then set one rule: no generation runs without your approval. Use Draft mode to check timing, never the pack. If you want help setting up that pipeline, here is how I work.

Sources

  • OpenAI — Introducing dots, 29 Sep 2026 — openai.com
  • OpenAI — ChatGPT release notes (dots, 29 Sep; Try on, 1 Oct 2026) — help.openai.com
  • Runway — Changelog: OpenAI Dots x Runway, 1 Oct 2026; Seedance 2.5 Draft Mode in Agent, 2 Oct 2026 — runway.com
  • Runway — Runway for ChatGPT (setup page) — runway.com
  • Pope Leo XIV (@Pontifex) — post on X, 2 Oct 2026 — x.com
خبرهای هوش مصنوعی را در تلگرام دنبال کنخلاصهٔ خبرها و برداشتِ من از کارِ واقعی، در کانالِ Nasravi.com در تلگرام.
عضویت در کانال تلگرام

Written by Mohammad Nasravi — AI creative director, in-house at a major FMCG group. Four years putting AI-made commercials past real brand teams, including Shirin Asal campaigns for Gulfood 2026 and Canton Fair 2025. More about the work · Start a project نوشتهٔ محمد نصراوی — کارگردان خلاق هوش مصنوعی، درون یک گروه بزرگ FMCG. چهار سال است کارِ ساخته‌شده با هوش مصنوعی را از فیلتر تیم‌های برند رد می‌کند؛ از جمله کمپین‌های شیرین عسل برای گلفود ۲۰۲۶ و کانتون فر ۲۰۲۵. دربارهٔ کارها · شروع پروژه بقلم محمد نصراوي — مدير إبداعي للذكاء الاصطناعي من داخل مجموعة FMCG كبرى. أربع سنوات من تمرير إعلانات مصنوعة بالذكاء الاصطناعي عبر فرق علامات حقيقية، منها حملات Shirin Asal في جلفود 2026 ومعرض كانتون 2025. المزيد عن العمل · ابدأ مشروعك