Daily Tech Update

The Report Local Post Insights method of the Google Business Profile Locations API has been deprecated. Google discontinued post insights on February 20, 2023, with no replacement. The method returns LOCAL POST INSIGHTSUNAVAILABLE . For...

The Create Card Token method of the Card Tokens API has been deprecated.

The following methods of the Transactions API have been deprecated: Create Transaction Get Transaction Capture Transaction Void Or Refund Transaction.

The Blog Posts and Stats API now includes a View Post method. Use it to increment a published post's view count when a site visitor views the post, and get the post's updated total number of views.

The Restaurants Reservations API now includes a Preview Cancellation Fee method. Use it to check whether a cancellation or no-show fee applies to a reservation and how much it would be, without cancelling the reservation.

The GSC Site Readiness API now includes 5 new methods: List Sites : List the connected Google account's Search Console properties. Get Site : Retrieve a property and its permission level. Get Sitemap : Retrieve a submitted sitemap and its...

The Automation Email Action API now includes 2 new methods: Assign Sender Details : Assign sender details to a send email action, so the action sends its emails from a specific name and verified address. Get Assigned Sender Details...

Apps can now expose tools and skills to AI agents such as Aria working on a Wix site, through 2 new APIs: Tools Host : Retrieve the tools of every installed app, run them, and switch an app's tools off for a site. App Skills : Retrieve the...

The Blog Settings API lets you manage the label of a site's all-posts feed and the settings that control comments. You can translate the feed label into the site's secondary languages, turn comments on or off, allow comments from visitors...

The Business Email APIs let you work with the Google Workspace products a Wix account purchases for its domains, such as Business Email. They include: Google Subscription : Check what an account can purchase for a domain, and read a...

The Feedback APIs let you collect and work through client feedback on a Wix site: Feedback Sessions : Open a round of review on a site and get a link to share with reviewers. Comments : Read, triage, reply to, and delete the comments...

An email design theme is a reusable definition of an email's look and feel, such as its colors, fonts, and spacing, that Wix users apply to Wix Email Marketing campaigns and automated emails. The Email Design Themes API lets you manage...

The Audience Service API resolves audiences, groups of participants defined by a set of criteria, into the individual participants they represent. With it, you can: Discover which audience providers are available. Resolve audiences into...

The Standalone Forms API lets you host a Wix form at its own public URL, so responders can fill it out without visiting a Wix site. Mark an existing form as standalone, query the standalone forms a site owns, and retrieve the site's...

The Wix WebMCP lets AI agents running in a site visitor's browser work with a Wix site through structured tools instead of reading the page and clicking through it. Site builders add the Wix WebMCP app, and visitor agents can then search...

When you upload a static site to a Wix-managed headless project, the upload now accepts video, audio, PDF, CSV, YAML, GeoJSON, 3D model, and pass-through image files, alongside the markup, style, script, image, font, and data files it...

Claude Haiku 5.5 is now on Amazon Bedrock and Claude Platform on AWS, positioned for subagents and high-volume work at roughly 75 percent lower cost than Haiku 4.5.

Google is turning Gemini into an agent that plans tasks, delegates to subagents, mixes models and gets its own workplace identity with an email address, starting in Gemini Enterprise.

A new llm plugin adds support for OpenAI's Decisions API, giving a quick way to try the API from the command line.

Z.ai's 753B-parameter mixture-of-experts GLM 5.3 coding and agent model is available on Bedrock through OpenAI-compatible APIs with prompt caching.

The llm-mistral plugin now supports reasoning models, which lets you call the newly released Mistral Large 4 from the llm CLI.

Anthropic published a new Usage Policy, so teams building client products on Claude should check what changed.

Falcon ASR is a newly released speech recognition model on Hugging Face that can be tested for transcription features.

Google's improved SynthID website is now available globally and can detect AI-generated content from Google, OpenAI and other providers.

Google AI Edge Foresight is a free experimental macOS app that transcribes meetings and audio fully offline using the on-device EmbeddingGemma 2 model.

NVIDIA describes fine-tuning a single Nemotron model family to reach gold-level results on both the IOI and IMO competitions.

A pattern for moving LLM API calls out of the Django request cycle into Celery tasks, covering job models, retries and status visibility to avoid timeouts and worker saturation.

Goodfire launched monitors that read a model's internal state to flag misbehaving agents and only call a second AI when something looks wrong, aiming to cut agent oversight costs.

Swift 6.4 ships with Swift Build as the default SPM build system, Subprocess reaches 1.0, and JavaScriptKit gets up to 40x faster WebAssembly bridging.

Safari Technology Preview 254 adds CSS size shorthands, border-radius side shorthands, interpolate-size and symbols(), so you can test them in WebKit now.

PostgreSQL removed the long-deprecated arraynulls setting on master for version 20 after finding pgdump never pinned it, which could silently turn null array elements into strings on restore.

An independent pgbench and TPCC-style benchmark finds PostgreSQL 19 Beta 3 and 4 performing about on par with 18.6 on OLTP workloads despite heavy code churn.

GitHub launched ReviewBench, an open benchmark built from real pull requests for comparing AI code review agents.

A new aws-ai-ml skill in the Agent Toolkit for AWS lets coding agents such as Claude Code, Kiro and Codex handle SageMaker inference optimization and benchmarking.

A worked example shows how an N+1 query pattern turns a 40ms database query into an 8-second API response, and how batching with a single IN query fixes it.

Amazon Bedrock AgentCore payments lets AI agents pay for services per request over x402, with spending limits enforced by the infrastructure.