OpenAI readies ChatGPT ‘superapp’ pivot ↗
OpenAI is reportedly turning ChatGPT from a Q&A chatbot into a broader “superapp” - coding tools, agents, image generation, and partner apps all gathered in one place. Almost disarmingly simple as an idea, but a huge product swing. (Fortune)
The larger bet is that users won’t just ask AI for answers. They’ll hand it unwieldy tasks - bookings, calendars, code, workflows - and expect it to do the thing properly. That’s the moat hunt now.
Perplexity lets AI models write their own search pipelines ↗
Perplexity’s “Search as Code” flips search from fixed API calls into model-written Python workflows. Instead of prodding a black box with query after query, the model builds its own tiny search machine. (The Decoder)
The company says it used far fewer tokens on a cybersecurity benchmark and led on most of its tests. Take the benchmark boast with salt, of course - but the idea is sharp.
Nvidia strikes fresh AI infrastructure deals in South Korea ↗
Nvidia announced a bundle of South Korea partnerships covering memory chips, AI cloud, data centres, robotics, and manufacturing. SK Hynix, SK Telecom, Naver, Doosan, LG Group, and Hyundai all show up in the machinery pile. (Reuters)
The big practical bit: Nvidia wants advanced memory supply locked down while pushing “AI factories” deeper into the industrial stack. Glamorous? Not especially. Important? Very.
The business of AI is facing 4 harsh realities ↗
The AI market story got a colder read: high costs, slower payback, infrastructure demand that is strong but not magically infinite, and financing that may stay pricey. That’s the wet cardboard under the gold-plated rocket. (Axios)
The key tension is simple: AI as technology can be wildly promising while AI as a business can still look expensive, lumpy, and under-monetised. Both things can be true - annoying, but true.
Banks lay groundwork for mass workforce cuts as AI takes hold ↗
Banks are reportedly shrinking junior analyst classes, in some cases dramatically, while still relying on those early-career pools for future AI talent. That’s a peculiar little ouroboros, biting its own graduate scheme. (Fortune)
The near-term AI use cases are more targeted than sci-fi: customer service, transaction monitoring, trade monitoring. Less “robot bank,” more “many small cuts and automations.”
Inside Amazon’s busiest European warehouse, where robots, lasers and humans deliver the future ↗
Amazon’s LCY3 warehouse in Dartford is running with mobile robots, AI software, scanners, and conveyor systems moving millions of units per week. It sounds half fulfilment centre, half industrial pinball table. (euronews)
The AI bit is practical: robot coordination, route optimisation, package measurement, label reading, and lane sorting. Not flashy chatbot stuff - more like the invisible skeleton of retail getting quicker.
FAQ
What does OpenAI’s reported ChatGPT superapp pivot mean?
OpenAI is reportedly pushing ChatGPT beyond a simple Q&A chatbot into a broader, app-like hub. The idea is to bring coding tools, AI agents, image generation, and partner apps into one place. Rather than only answering questions, ChatGPT would help complete larger tasks such as bookings, calendars, code, and workflows.
Why are AI agents becoming such a big focus in AI news?
AI agents matter because they move the product goal from giving answers to taking action. In this article, the central theme is users handing AI complex tasks and expecting it to complete them properly. That makes reliability, workflow handling, and valuable integrations more important than simply producing a polished response.
What is Perplexity’s “Search as Code” approach?
Perplexity’s “Search as Code” lets AI models write Python workflows for search instead of only calling fixed search APIs. That means the model can build a small custom search process for the task. The article says Perplexity claims this used fewer tokens on a cybersecurity benchmark, though the benchmark claims should be treated cautiously.
Why are Nvidia’s South Korea AI infrastructure deals important?
Nvidia’s South Korea deals matter because they connect AI infrastructure with memory chips, cloud, data centres, robotics, and manufacturing. Companies mentioned include SK Hynix, SK Telecom, Naver, Doosan, LG Group, and Hyundai. The practical goal is to secure advanced memory supply while pushing “AI factories” deeper into industrial operations.
What are the harsh business realities facing AI companies?
The article highlights four broad pressures around the AI business: high costs, slower payback, finite infrastructure demand, and expensive financing. The point is not that AI is failing as a technology. It is that building profitable AI businesses can still be costly, uneven, and harder to monetise than inflated expectations suggest.
How is AI affecting jobs in banks and finance?
Banks are reportedly reducing junior analyst classes while still depending on early-career workers as a future AI talent pool. The near-term uses described are practical and targeted, including customer service, transaction monitoring, and trade monitoring. Rather than replacing entire banks, AI appears to be creating many smaller automations across existing workflows.