Jaime Vengoechea
AI-powered. Shaped by Jaime Vengoechea.

The Applied Future.

AI, work, technology, and what comes next.

Sep 29, 2026Source-linked briefingPrevious editions RSS feed
In this edition

Five September 28 developments connect everyday AI work, vertical-video advertising and enterprise tools with stronger safety processes. Performance figures remain attributed to their publishers, and proposed safeguards are not treated as independently verified protection.

This Edition

Sep 29, 2026 · 5 stories
Google Display and Video 360 announcement artwork for Vertical Video Unification
Image: Google. Google Display & Video 360 is a trademark of Google LLC; no affiliation or endorsement.. Image source & credit.
Today's lead
MarketingSource: Google
Product announcement and advertiser test

Google brings vertical-video buying and measurement into one workflow

Google introduced Vertical Video Unification in Display & Video 360 to centralize programmatic buying and measurement across vertical-video publishers. It reports that an initial Hellmann's test with WPP Media increased unique reach by 24% while reducing cost per unique user by 25%. These are results from a named advertiser test, not a guaranteed improvement for every campaign or proof of increased sales.

Read original source
Context & source check

For a campaign test, compare incremental reach, frequency and business outcomes against the existing buying approach. A lower cost per unique user is useful only if those additional people matter to the campaign.

Product announcement and advertiser test. Company source. Source checked Sep 29, 2026. Not independently tested.

The reading list

4 stories

More from This Edition

4 stories
Latest edition
AI developmentsSource: Amazon Web Services
Model availability and vendor guidance

Sonnet 5.5 arrives on AWS for well-scoped coding and knowledge tasks

AWS announced Claude Sonnet 5.5 on Amazon Bedrock and Claude Platform on AWS. Its guidance positions Sonnet for bounded coding, document and spreadsheet tasks, while reserving harder judgment calls for Opus 5.5. AWS describes lower cost per task and faster execution for much routine work; those claims need testing on real workloads. Bedrock availability uses the Global CRIS inference profile, so teams should check their specific regional requirements.

Read original source
Context & source check

Route a small set of repeatable tasks to the new model and measure completed-task quality, latency and total cost including retries. Keep high-impact judgment and release decisions behind stronger review.

Model availability and vendor guidance. Company source. Source checked Sep 29, 2026. Not independently tested.

OpenAI and Basis customer-story preview with both company wordmarks
Image: OpenAI / Basis; marks belong to their respective owners. Image source & credit.
Latest edition
Work & wealthSource: OpenAI
Customer-reported internal evaluation

Basis reports faster tax-workbook completion in an internal AI test

An OpenAI customer story says Basis completed a complex 50-tab tax workbook in half the time with GPT-6 Astra compared with GPT-5.6 Sol. Basis also reports roughly 20% higher internal evaluation scores. Its checks include template compliance, primary-source consultation and reviewing final work. The article presents a customer test, not an independently audited benchmark or evidence that accounting review is unnecessary.

Read original source
Context & source check

Evaluate the completed work product, not just response speed. For a bounded internal trial, check template adherence, source traceability, assumptions and correction time before trusting a claimed efficiency gain.

Customer-reported internal evaluation. Company source. Source checked Sep 29, 2026. Not independently tested.

Latest edition
TechnologySource: Meta
Enterprise initiative announcement

Meta sets out an enterprise platform built around its AI stack

Meta announced an Enterprise Platform initiative focused on bringing the Muse agent, Meta Business Agent, Muse API and Muse Code to businesses and developers. The announcement appoints Chirantan Desai to lead the effort. It describes a strategic product direction rather than a complete procurement specification: this post does not establish pricing, deployment options or service guarantees for every component.

Read original source
Context & source check

Treat the announcement as a reason to watch documentation, not a reason to migrate. Before a pilot, require clear data handling, permissions, auditability, support and cost terms for the particular product.

Enterprise initiative announcement. Company source. Source checked Sep 29, 2026. Not independently tested.

OpenAI title artwork reading Towards safety cases for frontier AI training
Image: OpenAI. Image source & credit.
Latest edition
AI developmentsSource: OpenAI
Developing safety framework

OpenAI outlines stronger evidence and stop controls for frontier training

OpenAI proposed initial guidelines for safety cases before continuing frontier reinforcement-learning runs. They cover alignment evaluation, containment, monitoring, independent internal challenge, approval authority and incident investigation. The company says the recommendations are being implemented and will evolve. The post is a statement of developing practices, not proof that all controls are already operational or that deployment risks are resolved.

Read original source
Context & source check

The transferable operational lesson is to define stop conditions, accountable owners and retained evidence before increasing an agent's autonomy. A safety checklist is meaningful only when controls and incident response are tested.

Developing safety framework. Company source. Source checked Sep 29, 2026. Not independently tested.

Story preview