Asia Tech Times — Technology, science and culture across Asia

Technology, science and culture across Asia

Tech

Ant Group releases LingGuang 2.0, China's answer to Meta's Muse

Ant's October 27 upgrade turns LingGuang from an assistant you ask into an agent that watches, remembers and acts — free, where Meta's Muse charges $20 or $100 a month.

Comparison of LingGuang 2.0 and Meta's Muse across pitch, memory and action, security, connected services, pricing and availability.

LingGuang 2.0 and Meta's Muse at a glance.

From chat toy to life manager

LingGuang launched on November 18, 2025 as a multimodal assistant that answers through code — 3D models, charts, interactive maps and "Flash Apps", working mini-apps generated from a single sentence in about 30 seconds. Within two weeks users were already creating millions of them. Version 2.0 keeps all of that, and adds the missing dimension: time. Its errands run on an independent cloud host, so a task keeps updating after you close the app — for hours, days or weeks.

The core change is where AI steps in. Instead of waiting for a request, the agent — once asked and authorised — reads the services it is connected to, spots needs on its own, starts monitoring and follow-ups, and surfaces an alert only when the user must decide. Early hands-ons show the texture: concert-ticket on-sale times turned into to-dos, a gold-price tracker that fired on a 2 per cent drop, a family trip that flagged a ¥500-per-person fare drop and re-checked hotels overnight.

What the app actually shows

The shipping build centres on the Help Me Manage tab. A "Handed to LingGuang" feed shows each errand's live status and latest one-line briefing; a "More life ideas" shelf suggests things the agent can take over next — trip prep, fuel-and-charging logs, a weekly craft idea. The App Store subtitles the app "the personal assistant that gets things done".

The Help Me Manage tab: one errand in progress, a Shanghai-Guangzhou train-ticket watch, plus idea cards the agent offers to take over.

The Help Me Manage tab. (AI illustration, not an official screenshot)

Train-ticket watch. Holiday train seats sell out in waves, so the agent keeps a running brief: fare and seat status per train, which data source was queried, and the next scheduled check — then pings the moment seats appear. The same loop covers concert on-sales and airfare dips.

A running brief for a train-ticket watch: fares and seat status per train, data-source notes, and the next scheduled check.

Train-ticket watch. (AI illustration)

Group-dinner venue. One sentence — six people, Saturday, a hotpot place everyone can reach — yields a compromise-location map and a shortlist ranked by rating, price and distance, ready to share into the group chat. Because 2.0 remembers preferences, next time needs no re-briefing.

A group-dinner venue shortlist plotted on a map, ranked by rating, price and distance.

Group-dinner venue. (AI illustration)

Cinema and showtimes. For tonight's movie the agent sorts nearby cinemas by price or distance, flags the option that wins on both, and — once you pick — volunteers the detail humans forget: a departure reminder timed to traffic. Small task, but it is the 2.0 pitch in miniature: the AI does the noticing, you do the deciding.

Cinema listings for tonight sorted by price and distance, with the closest-and-cheapest option flagged and a departure reminder set.

Cinema and showtimes. (AI illustration)

LingGuang 2.0 vs. Meta Muse

Meta launched Muse in the US on September 8, 2026, and briefly topped the App Store free chart. The two products converge on the same thesis — memory plus action between conversations — and diverge on ecosystem and monetisation. The graphic at the top of this piece sets the two side by side.

Why it matters

Several big Chinese internet companies are racing down the same road, making personal agents the consensus bet of autumn 2026 rather than one company's moonshot. The contested ground is shifting from model quality to trust and ecosystem integration: whose agent can plug into the services where life actually happens, hold permissions safely, and finish what it starts. Ant's bet is that agent-mediated payments and service fees inside Alipay's loop matter more than subscription revenue — which is why LingGuang is free where Muse charges $20 or $100 a month.

The fine print

No public benchmarks exist yet for this product class — task-completion rates, failure explainability and long-horizon reliability remain unmeasured everywhere, Meta included. Permissions are per-service and revocable at any time, but the real test is behavioural: whether users feel comfortable enough to keep handing over errands after the novelty fades.

Availability

LingGuang 2.0 rolled out free on October 27, 2026, in Chinese iOS and Android stores and on the web at www.lingguang.com. Proactive management activates only for the services and data a user explicitly connects; chat, Flash Apps and the multimodal tools from 1.0 all remain.

Note on images: the four phone screens in this explainer are AI-generated illustrations of LingGuang 2.0's documented feature flows, recreated in English for readability; they are not official screenshots. The comparison graphic is an original rendering of reported facts.

Ask about this ChatGPT Claude Perplexity