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

Why OpenAI moved Habitat storage back to shared tenancy

OpenAI says Habitat returned to shared Cosmos DB storage after dedicated containers made product launches hard to provision. Hot-key limits, request priority and caching carry the tradeoffs.

The big change

What changed: OpenAI says Habitat now uses a shared Cosmos DB container by default, backed by hot-key throttling, lower-priority background requests and a serverless mode developed with Azure. Its earlier shift toward dedicated containers had reduced the blast radius of incidents but increased planning work.

Why it matters: The controls target different causes of interference: an exceptionally busy key, a background job competing with interactive traffic, and physical partitions losing per-partition throughput as storage grows. Together they explain how OpenAI says it reduced its need for per-team provisioning.

What to watch: OpenAI reports no provisioning or noisy-neighbor incidents in the months since its change, but offers no independent reliability study or public recipe for reproducing the custom service arrangement. Global placement, cache freshness and change-feed operations still impose tradeoffs.

Read the full story
—
Back to the briefing
guide

Put a complete process diagram in a Claude Code skill and check it

Map your owned procedure into a Claude Code skill with a Mermaid flowchart, check every branch in fresh sessions, and compare exact instruction tokens if API access permits.

The big change

What changed: Claude Code project skills can hold a workflow with a Mermaid diagram and adjacent instructions, so the diagram can map decisions while prose keeps precise operating rules.

Why it matters: A rendered diagram establishes syntax; branch mapping and matched task checks establish whether the skill still directs the work you own.

What to watch: Count the complete files under the same model and request shape. A lower count for that request is not a measured Claude Code session, billing or quality improvement.

Read the full story
—
Back to the briefing
analysis

Language-guided proton planning and its limits

How a research prototype turns instructions into planning adjustments.

The big change

What changed: The proposed language model interprets requests against prior computed plans, making its proposed trade-offs available for human inspection.

Why it matters: Choosing an optimization objective changes what a planning system tries to achieve. The weight assigned to an objective determines how strongly it competes with other goals, so accepting a request is only the start of evaluating its consequences.

What to watch: A dose distribution calculated for the planned conditions can change when patient positioning or proton range differs. Evaluating those uncertainties is a separate requirement from finding a favorable nominal result.

Read the full story
—
Back to the briefing
guide

Chain Claude Code skills into one reviewed GitHub issue

Use three Claude Code project skills to prepare an evidence-linked GitHub issue, secure exact human approval, then create and reconcile one issue with a checked readback.

The big change

What changed: Project skills can divide a business request into explicit evidence, drafting and review steps. Exact file hashes bind their handoffs to one approved brief; a separate handler controls the external issue write.

Why it matters: A maintainer can inspect each claim against repository records and approve the precise text before an issue can notify collaborators or trigger repository automation. A model-produced review sheet cannot stand in for that decision.

What to watch: The important boundary is the uncertain create response. The handler records intent before posting and reconciles the repository's full issue list against the same work key, title and body. A missing, duplicated or altered match requires human investigation before any further write.

Read the full story
—
Back to the briefing
analysis

OlmoEarth v1.3 offers annual satellite features for mapping

Ai2's annual embeddings supply mapping features that require task validation.

The big change

What changed: Ai2's satellite features can be reused across mapping tasks. Its earlier Studio route required a task model trained and run over imagery. The archive gives analysts precomputed vectors for mapping, but no land-cover labels.

Why it matters: A useful map still depends on reliable labels that define its classes and independent reference observations that show where predictions fail. Testing against training points alone cannot establish performance elsewhere. Analysts need to measure errors for the location and classes they intend to map.

What to watch: Differences between annual vectors can direct an analyst to places worth inspecting. The archive's provenance records identify contributing satellite scenes, providing a route back to the imagery. Before calling a difference land change, examine that evidence and measure classification and exact-year change accuracy against independent references.

Read the full story
—
Back to the briefing
guide

Route agent tasks by case features and audit accepted work

Build a deterministic per-case Agents SDK model route from owner-approved features, then compare every attempt and human-accepted task with a fixed baseline.

The big change

What changed: The application chooses a model for each approved case from three owner-labeled features: complexity, quality sensitivity and measured context size. A frozen rule puts the selected tier's exact model ID on that case's SDK Agent.

Why it matters: The choice can be checked against a fixed baseline using accepted work and full workflow cost, including failed attempts and review.

What to watch: This is a proposed reader-run method. Model quality, account charges and any saving remain unmeasured until the owner completes the trial and reconciles billing.

Read the full story
—
Back to the briefing
guide

Share one brand.md across two Claude Code skills

Create one approved brand.md, point two distinct Claude Code skills to it, then check a rule change against four fresh task runs.

The big change

What changed: Both manually invoked skills read one approved brand.md at an anchored relative path and report the file identity they used.

Why it matters: The owner can compare a baseline and a revised result for each separate task. A matching file hash establishes which bytes were read; the output still needs a factual and editorial check.

What to watch: A same-named skill from another scope can run instead of the printed project skill. Check the loaded skill's origin separately from the brand file's identity and each task's behavior.

Read the full story
—
Back to the briefing
guide

Use Claude Dispatch from your phone to check a Windows Notepad action

Send one bounded Notepad request from an eligible Claude Dispatch account, then inspect the exact local result and recover safely if the route or outcome differs.

The big change

What changed: Dispatch can hand a phone request to an app on the user's awake computer.

Why it matters: The user can request a local action while away from the keyboard. The actual result still needs a check at the computer.

What to watch: The useful boundary is the permission and result check. Keep the first action narrow, inspect the app afterward and stop when the session takes an unexpected route or its state is uncertain.

Read the full story
—
Back to the briefing