Five things worth a second look this week.

01 · AI + WORK

Barclays is cutting 5,000 jobs. Its chief executive says AI is only the icing.

On 22 October, Barclays said it would cut about 5,000 roles by the end of the year. That is roughly 5% of a workforce of about 93,000. About 1,000 had already gone in the first half of 2026. Most of the rest are operational and functional roles, and most of the savings come from the UK retail bank.

The same day, it reported pre-tax profit of £2.3bn for the third quarter, up 12% on last year.

The headlines tied the cuts to AI. The chief executive, C.S. Venkatakrishnan, did not. He said the reductions were "a result of the investments which we've made into technology", meaning modernised platforms, simpler systems and better processes. AI, he said, was "the icing on the cake".

Now put that next to what Barclays has said about AI elsewhere. On 1 October, Anthropic published a case study on the bank. A staff assistant built on Claude has been adopted by 16,000 colleagues. Claude models sort and route around 120,000 client emails a day in Global Markets. Barclays expects half its developers to be using Claude Code by the end of 2026. That page says nothing about jobs.

So a bank is rolling AI out across thousands of staff, cutting 5,000 roles in the same year, and telling us the two are not the main story.

Both things can be true. Modernising old systems and adding AI are hard to separate inside a bank, and nobody outside it can see which saving belongs to which. Not "AI took the jobs". Not "AI had nothing to do with it". The company that knows the split is also the one choosing the words. None of the reporting gives that split.

02 · AI + FILM

China's first AI-made film to reach cinemas has made about $180,000 against a $1.5m budget.

Sanxingdui: Future Past is a sci-fi thriller about an archaeologist trying to save a future world enslaved by AI. It comes from Bona Film Group. The $1.5m budget is AFP's figure. The box office, about $180,000 through Monday, is from the ticketing platform Maoyan, via Fortune.

Bona's chief operating officer, Jiang Defu, called it an important experiment. The USC professor Stanley Rosen told Fortune he doubts people will pay for an AI film in cinemas unless it matches the best of Chinese blockbusters, and suggested streaming may suit it better.

The odd part is where AI already won. According to Caixin, over 95% of the more than 120,000 microdramas released in China in the first quarter of 2026 were AI-generated, up from almost none a year earlier. Same technology, different screen. Whether the cinema flop is about AI or about the film is not something one weekend can tell us.

03 · SOCIAL + VIDEO

Instagram wants your long videos, and it has told you what it calls them.

On 26 October, Instagram's head Adam Mosseri and vice president Tessa Lyons announced a test of long-form horizontal video called Features, in a six-minute video. The idea is that creators use Reels to build an audience for the longer videos. The test starts in the US only, according to The Hollywood Reporter. Mosseri said it is early and they are testing and listening to feedback.

Reports put the length limit at 90 minutes. The announcement coverage does not confirm that, and it does not say where Features will sit in the app or whether creators will share ad revenue.

04 · CREATOR ECONOMY

Meta is reportedly paying YouTubers $50,000 to $500,000 to post on Instagram.

Business Insider reported on 29 October that Meta has offered some YouTubers deals starting at $50,000, with some proposals reaching $500,000, to upload long videos to Features. The deals reportedly aren't exclusive, so creators can keep posting to YouTube. One unnamed source said the main condition is posting to Instagram the same day as YouTube.

Meta confirmed to Business Insider that it funds some creators directly, and declined to give contract details. It did not deny the reports. No creators are named. It is not known whether the money is per video, monthly or one-off.

Mosseri said on 28 October that Instagram is exploring ways to pay long-form creators. For now, the payment looks like a deal and not a system.

05 · AI + SECURITY

Your AI agent is about to ask for your passwords. 1Password wants it to ask for permission instead.

On 22 October, 1Password announced 1Password for Agents. The agent proves its identity with a digital key. You see which login it needs and why, and approve or decline on your own device. 1Password then fills the login into the sign-in form itself. The agent carries on with the task, and the password never reaches it.

Named agents include Anthropic's Claude Duet, OpenAI's ChatGPT Work and dots, Meta's Muse and Grok Bot. The release says 1Password is "building with" them and that more integrations are to come. An Anthropic executive says Duet "asks before it signs in anywhere", and that "soon" 1Password will complete the sign-in. So it is not clear which of these work today. Pricing isn't given.

The question under it: when software acts as you online, who holds the keys? This is one answer. It is not the only one.