日本語
2026-07-23 Evening edition
AI News Daily

AI News Daily
2026-07-23

Evening edition

Cheaper, leaner models from Google; Anthropic turning its own usage data into a product surface; and more than $5 billion of US federal money moving into AI for science.

02

Today's highlights

03
Company news

Anthropic releases an Economic Index connector for Claude

Published: 2026-07-22
  • Announced 22 July 2026: Claude can reference Anthropic Economic Index data directly.
  • Turned on from the connector menu in Claude.ai, with no additional installation, and it works on any Claude model.
  • Users can ask in chat how AI is used by occupation and how usage trends differ by region.
Why it matters

Companies can see, from data, how AI is actually used in work, making it easier to pick priority areas for adoption and to design internal enablement.

Extra setup
Zero
No additional installation, and it runs on any Claude model.
Source: https://www.anthropic.com/news/anthropic-economic-index-connector (Tier 1, Anthropic official website)
04
Model release

Google announces Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber

Published: 2026-07-21
  • Three models announced on 21 July 2026, including 3.5 Flash Cyber, which specialises in vulnerability detection.
  • 3.6 Flash consumes 17% fewer output tokens than 3.5 Flash, while its DeepSWE coding score rises from 37% to 49%.
  • Output pricing falls from $9.00 to $7.50 per million tokens.
Why it matters

Stronger performance in the low-cost tier means business applications may raise response quality while holding existing API costs down.

DeepSWE score
49%
Gemini 3.6 Flash, up from 37% for 3.5 Flash.
Source: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/ (Tier 1, Google official website)
05
Regulation and policy

White House puts more than $5 billion behind the Genesis Mission

Published: 2026-07-22
  • On 22 July 2026 the White House Office of Science and Technology Policy announced more than $5 billion in federal funding to expand Genesis Mission, a national programme applying AI to scientific research.
  • More than 15 federal agencies supply research grants along with data and computing infrastructure.
  • 278 research projects were selected from over 5,000 applications.
Why it matters

Large government funding for AI research is a leading indicator of which R&D fields adopt AI next and of the shared infrastructure built to support them.

Projects selected
278
Chosen from over 5,000 applications.
Source: https://www.whitehouse.gov/releases/2026/07/45502/ (Tier 1, The White House official website)
06

Trend overview

Two tracks at once

The major players keep investing on both fronts: efficiency in foundation models, as in Google's refresh of the lightweight Flash line, and edge and physical AI, as with NVIDIA Cosmos.

Data as differentiation

Anthropic is not competing on model performance alone. It is using its own data to ship business-facing features, an angle rivals cannot copy with a benchmark score.

From rules to funding

The US government is strengthening its push to lead not only on AI regulation but on investment in the use of AI for scientific research.

07

Wrap-up

What to watch next

Whether price and efficiency gains in the lightweight tier hold as workloads scale, whether vendor-owned datasets become standard inside assistant products, and what the 278 funded Genesis Mission projects deliver.