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 38 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.

guide

Calibrate a judge for a Claude Code Skill revision

Calibrate a source-grounded second-model judge for each criterion, then compare one owned text-output Claude Code Skill revision with its baseline on real development and sealed held-out cases.

The big change

What changes: A source-grounded second model can assess a criterion that literal string checks cannot, with its own calibration receipts and an explicit unknown result when evidence is thin.

How it works: The same frozen input, truth, tests, model controls and one-attempt rule bind both Skill revisions. Every calibration and grading call records its input and output hashes before a score is computed.

Why it matters: A passing text match cannot establish source fidelity. The owner can inspect the judged answer, disagreements and regressions before choosing accept, hold or revert; the decision does not activate the revision.

Read the full story
—
Back to the briefing
news

Springer Nature tests a private AI assistant for peer reviewers

Springer Nature's private peer review prototype has no reported results.

The big change

What changed: Springer Nature and Google DeepMind are testing an assistant that points reviewers toward parts of a manuscript needing closer attention, including text, charts and figures. That could make AI a guide to where a specialist looks. The reviewer still decides whether the method and conclusions withstand scrutiny.

Why it matters: A suggestion can shape the focus of a review even when the system does not write the final recommendation. Springer Nature's policy keeps evaluation and the report with the reviewer. Private hosting is the publisher's confidentiality approach. It says little about the quality of review suggestions, and the release gives no independent security or retention details.

What to watch: Results from this prototype need to show what concerns it finds and misses, and how much work reviewers spend checking unhelpful prompts. Comparing assisted and unassisted reviews would help separate quicker navigation from better scientific reasoning. Reviewers need access, retention and deletion details to judge how unpublished manuscripts are protected.

Read the full story
—
Back to the briefing
news

Australia’s cyber alerts rose sharply. Its AI attack count is unknown

ASD warnings rose almost 80%; its published report gives no AI-attributed attack count.

The big change

What changed: ASD's warnings rose almost 80% and incident responses rose 20%. They measure different agency actions; neither attributes the increase to AI.

Why it matters: ASD advises independently verifying unexpected requests for credentials, money or security changes through a trusted channel. That changes the reader's decision from whether the message sounds convincing to whether its claimed sender can be confirmed.

What to watch: A credible claim about AI's share would need a disclosed definition of AI involvement, case-level attribution and comparable incident totals for each period. Readers could then ask whether the same kinds of cases were counted in both years before deciding what a change means.

Read the full story
—
Back to the briefing
guide

Expand an OpenAI Agents SDK ticket pilot in reviewed phases

Copy a local ticket pilot that moves from observation to reviewed actions and one-effect tagging, with exact approvals, audits and no-replay recovery.

The big change

What changed: The pilot moves from reading one owned ticket to applying a narrow label through reviewed policy changes. The supervisor can propose more than its local policy will execute; that policy decides whether an effect lands.

Why it matters: An operator can compare a proposed action with the exact ticket and audit receipt before expanding permissions. A model answer cannot grant itself authority, and the bounded tier still permits at most one tag effect in a run.

What to watch: Promotion depends on the operator's review of real owned cases. This kit has no provider or helpdesk trial, and its files are editable; a shared service would need authenticated review and transactional storage before using this pattern.

Read the full story
—
Back to the briefing
guide

Build a landing page with a playable AI video and real 3D

Generate and inspect a Flow clip, add it to a copyable Vite and Three.js page, then check full playback and motion behavior at the hosted URL.

The big change

What changed: A creator can combine a generated, downloadable film with live WebGL geometry in one page. The film remains a video element; the shape is rendered by Three.js and responds to scroll on eligible screens.

Why it matters: A visitor can choose when to play the film and stop decorative motion. The same page still has readable copy when the browser or device cannot run the 3D effect.

What to watch: The decisive checks happen on the published URL. An MP4 header and a successful build do not establish that the hosted clip decodes, that a phone can play it, or that the destination is appropriate for the intended audience.

Read the full story
—
Back to the briefing
guide

Carry a reviewed project lesson into the next Claude Code run

Scan an owned project note, run a bounded Claude analysis with the pinned revision, review its source-linked proposal, and load the accepted lesson next time.

The big change

What changed: A scan fixes the source bytes, rights note and task for the next analysis. Source text is treated as data, so a sentence inside a note cannot authorize a new instruction or action.

Why it matters: The process writes a pending intent before a Claude call. Its exit produces a proposal, never an automatic memory update. A person checks the quote, inference and scope, then accepts or rejects the whole proposal hash.

What to watch: Acceptance creates a new immutable fact revision. The next context names that revision. A missing or uncertain result holds the store until the owner reconciles it; the wrapper does not silently replay a potentially billed call.

Read the full story
—
Back to the briefing
guide

Build and check a monthly workload model with ChatGPT for Excel

Use OpenAI's Excel add-in to create a monthly backlog model from your own approved counts, update one input or formula range, then reconcile the workbook before sharing.

The big change

What changed: ChatGPT can work in an open Excel workbook through its add-in, so a monthly model can be built and updated in the same file.

Why it matters: The owner can trace each month from approved counts to the next month's opening backlog, then check a revised version.

What to watch: An apparently correct total or green status cell cannot establish that every source value and formula is right. Inspect the workbook before sharing.

Read the full story
—
Back to the briefing
guide

Screen every candidate document before a Claude Code answer

Use two document scouts and a local checker to account for every file, approve an exact selection, then verify the answer against selected originals.

The big change

What changed: The reader gets an exact selection and a reason for each exclusion before the main agent opens original files.

Why it matters: A missing or changed candidate stops the handoff; a plausible scout report alone cannot silently narrow the pool.

What to watch: The human still judges relevance, source meaning and the final answer. The checker establishes file and decision consistency, not truth.

Read the full story
—
Back to the briefing