THE WORLD IS NOT STANDING STILL.RSS
BIG CHANGE.
BIG CHANGE, one tap away
BIG CHANGE, one tap away

Make room for the big picture.

Add BIG CHANGE to your home screen or desktop. It opens in its own window, with the latest stories when you’re online.

iPhone & iPad
Open bigchange.ai in Safari. Tap Share, then Add to Home Screen. Turn on Open as Web App if shown, then tap Add.
Android
Open bigchange.ai in Chrome. In the browser menu, choose Install app or Add to Home screen.
Desktop
In Chrome or Edge, use the install icon in the address bar or the browser’s app menu. In Safari on Mac, choose File → Add to Dock.
All briefings

Follow the change ·

The day,
in perspective.

What happened in AI and robotics — and the change behind it. A reading path through 50 stories from BIG CHANGE.

Ordered using RADAR attention, story opens and likes. Excerpts from our published reporting; open a story for the full context and sources. Dates in UTC.

analysis

DetectorScope exposes the features behind AI-text classification

DetectorScope shows contributing AI-text cues without proving authorship.

The big change

What changed: DetectorScope reveals the features behind its score, giving reviewers cues to challenge.

Why it matters: A 2025 evaluation of other detectors shows why the share of generated text caught must be read alongside the rate at which human text is flagged. An inspectable reason for one score cannot tell an editor whether that error rate holds for their submissions.

What to watch: Research on older detectors found higher false-positive rates for non-native English writing. Before treating any flag as evidence against a writer, an editor needs results for the writing population at hand and a way to challenge the particular cues.

Read the full story
—
Back to the briefing
analysis

Commerce.com adds merchant opt-out to WebMCP storefront tools

Commerce.com and Cloudflare's WebMCP bridge honors merchant opt-outs. Storefront MCP governs shopping tools; the buyer completes checkout.

The big change

What changed: Cloudflare's rule checks each merchant hostname before serving the page. A platform sharing one zone can deliver its WebMCP bridge across stores while withholding it from merchants who opt out. The previous zone-wide switch could not express that choice for one store.

Why it matters: The bridge exposes only tools the store enables. With Storefront MCP active, an agent can alter the shopper's cart; the buyer completes the purchase. That creates two merchant controls before checkout.

What to watch: Shopify's checkout tools allow submission after buyer confirmation and return control for 3D Secure authentication. Commerce.com's tools hand over a checkout link. Merchants should compare where each tool's authority ends.

Read the full story
—
Back to the briefing
analysis

Apollo’s AI payoff claim exceeds what its charts show

Company ratios, productivity and AI spending measure different outcomes. Connecting them requires evidence from the businesses adopting AI.

The big change

What changed: Apollo’s challenge to AI returns exposes a measurement gap: sales to adopters do not establish gains for those adopters.

Why it matters: A payment is revenue for one business and a cost for another. Neither side of that transaction establishes whether the purchaser’s work became more valuable after all costs.

What to watch: The useful evidence connects an adopter’s expenditure to its own outcomes over a defined period, with a credible comparison. That would let managers assess a return without treating an economy-wide average as their business case.

Read the full story
—
Back to the briefing
analysis

Liquid RIPL reports complete answers under 50 ms at peak load

Liquid reports fast complete answers. Comparisons change both model and engine.

The big change

What changed: Liquid is putting complete generated answers during sustained traffic at the center of its RIPL pitch. For an application answering a live request, the useful clock runs through the final output, not merely the first token. Retrieval, business logic and network delivery still add time outside that model measurement.

Why it matters: Liquid's published speed ratios compare different models as well as different serving engines. An operator cannot read those ratios as an engine-only gain. The deployment question is whether a complete configuration meets the task's answer-quality and tail-latency targets at the traffic rate it must handle.

What to watch: A comparable evaluation would use the same tasks, timing boundaries and request load, then check quality and cost beside latency. Holding the model fixed would better isolate RIPL's contribution. Measuring the entire application path would show whether a fast model answer fits the request's end-to-end budget.

Read the full story
—
Back to the briefing
analysis

Bayer's AI agents diagnose crop-research workflow failures

Bayer's workflow-diagnosis agents shorten reported support time; accuracy remains unaudited.

The big change

What changed: Staff can ask an agent about a research run. The diagnostic task shifts from finding records in separate systems to judging whether those records support the answer.

Why it matters: Faster access to evidence could reduce support time. The answer still needs to identify the run and explain what its records establish; a quick reply alone does not show that the right cause was found.

What to watch: For a research-operations team, the decision is whether each answer identifies the right run, cites observations sufficient for its conclusion and stays within the caller's permitted data. A completed agent request, a plausible diagnosis and a repaired workflow are different outcomes to measure.

Read the full story
—
Back to the briefing
guide

Carry a reviewed Stitch DESIGN.md into two checked Claude Code components

Export a real Stitch DESIGN.md, approve its exact revision, import it into Claude Code and check two components against the same design.

The big change

What changed: Stitch supports exporting design rules as DESIGN.md, which can travel with a project into a coding workflow.

Why it matters: One approved file can give two component tasks the same named visual rules, while code and screen reviews catch departures from them.

What to watch: The export, project import and generated components each need verification. A missing export stops this workflow; a generated screen alone does not establish working behavior.

Read the full story
—
Back to the briefing
guide

Organize two ongoing tasks with Claude Cowork Projects

Set up two Claude Cowork Projects for separate ongoing tasks, then check each project's instructions, references, account-memory choice and first output.

The big change

What changed: Cowork Projects can hold instructions, context and memory for related tasks. You can start fresh, import one existing Claude project, or connect a local folder.

Why it matters: A separate project for each ongoing task gives you a place to keep its working rules and references together. Checking a first output in each project shows whether those settings support your actual work.

What to watch: Project memory can still use account memory unless you turn that option off. Uploaded files and connected local files also have different storage and access requirements. Inspect both before relying on a project.

Read the full story
—
Back to the briefing
analysis

ThinkingBox Training improves repeated success as task coverage narrows

Repeatability rose; coverage narrowed.

The big change

What changed: Final-state checks reward agent episodes. A final-state check asks whether the database actually contains the intended result. That matters because a persuasive reply and a correct record can part company.

Why it matters: Repeated trials reveal a kind of reliability that one successful run cannot. An average success rate obscures whether an agent reaches many tasks occasionally or handles a narrower set consistently.

What to watch: Outcome checks are only as good as the task and policy requirements they encode. Engineers need repeated trials on authorized tasks that represent the intended workflow.

Read the full story
—
Back to the briefing