๐ง Research
Courses given an AI tutor saw lower grades in a 2,379-student trial ยท Measure it
A randomized trial at a large US university, covering 2,379 students in 30 instructors' courses, gave some sections a built-in AI tutor. Those sections finished about four points lower out of 100 and did even less of the online coursework, though only about 15% of their students used the tutor. The authors think it led to wider AI use that replaced coursework. It's a working paper, not peer reviewed, and it uses course grades rather than a separate test. One lesson carries over: when you add an AI helper, track what people stop doing.
๐ Launches
Synthesia launches in ChatGPT, on all its plans ยท Build it
Describe a video in ChatGPT, or share a doc, deck or spreadsheet, and Synthesia builds it in the chat, by default as a quick motion-graphics summary with b-roll. Say "make it in Studio" for an avatar that reads your script word for word, with scene-by-scene editing in Synthesia's editor. Synthesia says this mode suits training, onboarding and compliance. It's live on every Synthesia plan as of November 3, but the post doesn't say how credits work or which ChatGPT plan you need, so check before planning a batch.
Colossyan can now dub uploaded videos into 70+ languages ยท Roll it out
Upload an MP4 or MOV, pick the language, and Colossyan dubs it with the speaker's own voice or an AI voice, with the mouth movements synced. Videos can run up to 15 minutes for now and work best with one clearly visible face. On-screen text isn't translated yet. Dubbing is free for an initial period, and Colossyan hasn't said what it will cost later or which plans get it.
Go1 adds OpenAI Academy's AI courses for workplace teams ยท Roll it out
Starting November 4, organizations on Go1 can add three OpenAI Academy courses: AI Foundations, Applied AI Foundations, and Agents and Workflows. Admins assign them like any other Go1 course and track progress in Go1's reports. Your organization activates it through Go1, and the page points you to a demo, not a switch you flip yourself. It's new content, not an authoring tool, and Go1 gives no price, course length or badge details. The announcement is on its blog.
๐ ๏ธ Try this
Turn an SME recording into a short captioned clip with OpusClip 4.0 ยท Build it
OpusClip 4.0 takes raw talking-head footage, cleans up the speech, adds captions, and turns spoken numbers into on-screen graphics. You get 60 to 90 second clips or a 3 to 5 minute cut. The free plan gives one short watermarked clip a day, with three days to download it, and OpusClip's pages don't say which 4.0 features it includes. Untested here: it was built for social clips, so check that it kept the steps, not just the highlights, and have the SME review it before anyone learns from it.
โก Quick links
HeyGen's API can edit a finished video scene by scene ยท Build it: Change a script line, voice or avatar style, and usually only the changed avatar scenes re-render and get billed. Scenes with no avatar are billed on every render. It's an API feature with no no-code route in the docs, dated "November 2026" with no exact day.
Tango's guides can answer questions inside Copilot and ChatGPT ยท Roll it out: Staff ask "Check our Tangos for [task]" and get your steps and screenshots back, read-only. Enterprise plan only.
AI for L&D is one short email each Sunday on what's new in AI for instructional designers and L&D teams. Know someone figuring out AI for training? Forward this. Got a tip? Just reply.