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 37 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

Anthropic proposes a 5% share allocation for a public benefit fund

Anthropic says a 5% share allocation awaits stockholder approval. Its board may add further equity, while fund governance and the valuation baseline remain undisclosed.

The big change

What changed: Anthropic has proposed reserving company equity for public-benefit uses through a wholly owned fund. The initial 5% allocation still needs stockholder approval, and shares do not themselves provide a stated cash budget.

Why it matters: Equity could give the fund a stake in value created by Anthropic’s AI business. How that value reaches outside recipients depends on decisions about the shares and awards that the announcement has yet to describe.

What to watch: Anthropic has promised details on governance and valuation milestones. Those disclosures should show who chooses recipients, what authority the fund has, and when the board can add further equity.

Read the full story
—
Back to the briefing
news

Philadelphia delivery robot arrest puts sidewalk rules in focus

Carolyn Marvin was arrested after blocking an Uber Eats robot in Philadelphia. Police and Marvin give separate accounts; Pennsylvania requires remote oversight of authorized delivery devices.

The big change

What changed: Avride is authorized to run delivery robots in Philadelphia in partnership with Uber Eats. Pennsylvania’s Phase 2 rules allow autonomous or remote movement, while requiring remote monitoring and the ability to take control when necessary. Robot delivery on a public route is a supervised commercial operation.

Why it matters: Sidewalk users and delivery workers now encounter that operation in shared public space. The state requires devices to yield to pedestrians in pedestrian areas; the arrest has drawn attention to how a protest against the service meets an active delivery route.

What to watch: Marvin says she will contest the disorderly conduct citation. Any court outcome would address her case. PennDOT’s operator authorization establishes the framework for the robots, but the public material reviewed does not show how the device or a remote operator responded during this encounter.

Read the full story
—
Back to the briefing
analysis

Liquid’s d1-ultra adds a hosted model for decisions without generated text

Liquid AI released d1-ultra as a hosted model that returns probabilities for structured text and image decisions without generating answer tokens. Its benchmarks and price come from the company.

The big change

What changed: Liquid now offers its larger decision model through an API. A team can ask several defined questions about the same text or image in one request and receive a probability for each answer, without paying for generated output tokens.

Why it matters: For developers building routing or inspection software, the useful output is often a label, score or yes/no estimate. The fixed answer types make that result directly available to downstream rules. Liquid's billing counts the input again for every question, so adding checks to one image also adds cost.

What to watch: The decision for an operator is where to set thresholds and when to send uncertain cases to a person. Independent benchmark results can inform model selection, but a score across general tasks cannot set a reliable threshold for a company's own images, labels or error costs.

Read the full story
—
Back to the briefing
analysis

Goodfire's neuron interpretability thesis faces a superposition objection

Nick Cammarata argues that many apparently mixed-purpose neurons can be studied directly. His Goodfire essay offers preliminary evidence, while Neel Nanda disputes the broader conclusion about superposition.

The big change

What changed: Cammarata offers a way to study some model behavior through the network's own neurons and circuits. His reinterpretation of earlier examples and preliminary language-model audit challenge one explanation for why neurons seem to mix meanings; they do not establish that the whole network is understood.

Why it matters: Researchers investigating a model's answer could use neuron-level traces without first training a separate model to extract features. That would make direct investigation more practical where the traces are reliable, but a readable trace can still omit important computation.

What to watch: The promised technical paper needs to show how the language-model audit was done and how its explanations hold up, especially for weak activations. That evidence will help determine where studying native neurons is useful and where other methods remain necessary.

Read the full story
—
Back to the briefing
analysis

HybridAgent tests latent reasoning alongside text-based tool use

A new preprint and released code show how HybridAgent alternates latent reasoning with text tool use. Its measured gains come from the authors' benchmark setup.

The big change

What changed: HybridAgent gives a language-model agent a way to move between internal continuous-state steps and ordinary text during one task. Text remains the interface for search, Python and the final answer.

Why it matters: Developers studying agents can examine whether every intermediate step needs to be written out. The authors' matched experiments suggest fewer generation steps for their method, while the public code makes its implementation inspectable. Neither establishes a lower bill for an operating service.

What to watch: Independent runs with the released checkpoints can test whether the measured advantage survives different tasks and tool costs. The harder inspection question is whether the readable words recovered from a latent state faithfully describe how an answer was reached.

Read the full story
—
Back to the briefing
guide

Create and check a complete scene in Google Flow

A documented route from permitted Google Flow clips through version choice and Scenebuilder assembly to a downloaded scene checked from beginning to end.

The big change

What changed: Flow's documented workflow reaches beyond generating a clip: it keeps edit versions, arranges clips in Scenebuilder and downloads the complete sequence as one file. Google does not date the introduction of these controls, so this is a description of the current capability.

Why it matters: A creator can choose a version for each shot, adjust the order and trims, then compare the downloaded scene with the intended sequence. That final file check catches problems an attractive individual clip cannot reveal, including a missing shot or awkward join. Google's editing guide documents those assembly controls.

What to watch: The practical decision turns on the account's available model and regional controls, the current credit quote, and whether the downloaded file passes a full image and required-audio check. Those checks determine whether to keep the scene or repair a specific clip or cut.

Read the full story
—
Back to the briefing
news

Alliance Arts cancels public Drunken Goddess Reflux demo for a closed playtest

Alliance Arts canceled its October 19 public demo, citing difficulty managing sharing restrictions and concern about pre-launch imitation using AI or other tools. A limited playtest is planned instead.

The big change

What changed: A demo that anyone could have accessed is becoming a lottery-based playtest with rules agreed before participation. Alliance Arts is narrowing access while still seeking player feedback.

Why it matters: The publisher's two stated concerns—managing gameplay sharing and possible imitation—now shape who can play before launch and how participants can share what they see. Players lose the planned public demo; selected applicants get access through a controlled process.

What to watch: The announced playtest runs October 23–31, subject to schedule changes. Its terms and Discord feedback process show how Alliance Arts intends to manage this test, but the announcement provides no evidence that the process will prevent copying or that copying has occurred.

Read the full story
—
Back to the briefing
guide

How to rewrite a selected Word passage with Claude and review changes

Select one passage in Word, ask Claude for a bounded rewrite in tracked changes mode, then inspect and decide each revision before saving.

The big change

What changed: Claude for Word can limit a requested edit to selected text and submit it as Word revisions. The document owner can see the text removed and added before making either change final.

Why it matters: A person polishing a report or memo can ask for clearer prose while retaining control over its facts, qualifications and voice. The review happens in the document where the passage will be used.

What to watch: Selection limits the edit, not the add-in's access to the open document. Permissions, retention policy and a careful read of every proposed change still determine whether this workflow suits the material.

Read the full story
—
Back to the briefing