Shift Lab

10/29/2026

Figma's October 22 Enterprise update is about control rather than canvas. Admins can set default paid AI credit access at the organization and billing group level, manage that access through SCIM, block company-email signups to external Starter and Professional plans, and reach teams, folders and files through new admin REST APIs.

That last piece matters to anyone running design at scale. File sprawl is a real cost, and an API is how you audit it instead of asking around.

Shift Lab works inside client design systems every week, and the hardest problems there are rarely visual. They are provenance questions. Which file is canonical, who is allowed to change it, what happens when a contractor's access ends.

Tooling that answers those questions is design system infrastructure, even when it looks like admin settings.

Read more: Figma release notes, October 2026 (https://releasebot.io/updates/figma)

Complete list of Figma latest updates for October 2026: get every product news, release note, and changelog from Figma summarized in one timeline.

10/28/2026

New research covering 718 firms and roughly 300 million work events found that AI coding agents produced about 30 percent more code but only 12 percent more resolved issues. Average review time per pull request rose 49 percent.

This is the failure mode Shift Lab designed around. Our delivery model moves repetitive work into a structured pipeline, but the review points are deliberate rather than residual. Per-story acceptance criteria, PR review, sprint review, production sign-off. Each one sits where it does because output without a gate simply relocates the problem.

Volume is not throughput. A team that generates faster than it can review has bought a longer queue and called it progress.

The useful question for any agency making AI claims right now is a simple one. Can they show you what happened after the code was written.

Read more: InfoQ, AI-generated code fails to boost productivity due to downstream bottlenecks (https://www.infoq.com/news/2026/10/jellyfish-ai-productivity/)

A recent Harvard study of 718 firms, using the Jellyfish analytics platform, finds that although AI coding assistants and agents increase code volume, this does not translate into a corresponding increase in completed work, partly because AI-generated code increases the workload in later stages of d...

10/27/2026

Shift Lab is not an AI agency. It is a tech-first digital agency that changed how it works.

The same disciplines run as always, product strategy, design, technology, project management and review, with the repetitive parts of each stage moved into a versioned pipeline of agent skills that every project fetches from one served system rather than keeping its own static copy. Human checkpoints stay deliberate: spec sign-off, design review, architecture review, PR review, production sign-off.

What makes it defensible is that it is instrumented. Across 407 pull requests between September 26 and October 26, org-wide, the first automated review came back clean on 82. About 20 percent, with a median of two iterations to clean.

10/26/2026

Digiday's 2026 agency report found 77 percent of respondents increasing AI search and GEO investment over the next year, with roughly 40 percent of it reallocated from general marketing budgets.

That budget is moving toward a problem that lives in the stack. Structured data, server rendering, content modeling in Contentful or Sanity, and parity between what a crawler receives and what a visitor sees.

This is the kind of work Shift Lab does alongside creative agencies. The partner agency owns the brand and the client relationship. We handle the technical layer underneath, often invisibly, so the strategy actually ships.

Creative teams are fielding technical questions from clients right now. Having a development partner already in the room is a faster answer than building that capability from scratch.

Read more: Digiday Media Agency Report 2026 (https://digiday.com/media-buying/digiday-media-agency-report-2026-how-agencies-adapt-to-a-media-landscape-shaped-by-ai-and-creators/)

Media agencies are navigating an increasingly complex advertising landscape — and advertisers are reassessing their media strategies.

10/22/2026

Agency pricing is splitting. Digiday reports some groups absorbing AI costs, some building tokens into their rates, and others moving clients to output-based fees. The common problem is proving that usage turned into value.

For partner agencies, that question lands on their technical partner too: what does the work cost now, and what did it buy?

Our answer is to measure the delivery process instead of arguing about it. Every story through our pipeline produces a structured record, so there is real data on review iterations, test results and acceptance-criteria checks behind how we work.

That gives creative agencies something concrete for their own client conversations. We stay an invisible technical partner, and the numbers stay honest.

Pricing models will keep shifting. Evidence travels better than rate cards.

Read more: Digiday, ad agencies and AI costs (https://digiday.com/media-buying/were-starting-to-wonder-ad-industry-chases-ai-value-as-usage-outpaces-proof/)

Every choice has a price and the bill always comes due. Marketers are only just starting to work out what theirs actually costs for AI.

10/21/2026

shadcn/ui's October update adds Trace and Code Block, plus 21 AI Elements patterns for agentic interfaces. Trace lists the steps an agent takes, with a status for each. Code Block handles logs, terminal output and diffs. Both are available for Base UI, React Aria and Radix.

Agent-facing UI is becoming a normal request on client roadmaps. The design question is what a running, failing or pending step should look like, and the answer needs to live in the system, not on one screen.

That is the work we do in component libraries: take primitives like these, apply a client's tokens, document them in Storybook, and test accessibility at the component level so every product built on top inherits it.

Copy-in components mean the library is yours. We help teams keep it coherent as it grows.

Read more: shadcn/ui changelog (https://ui.shadcn.com/docs/changelog)

Latest updates and announcements.

10/21/2026

GitHub made agentic autofix for CodeQL alerts generally available on October 15. Copilot proposes a fix, reruns CodeQL to validate it, and opens a draft pull request for review.

We like the shape of it, because it ends with a person. On our projects every pull request goes through automated review and then a human, and the same applies to anything an agent writes, security fixes included.

The practical gain for a client is a shorter path from alert to reviewed patch. Our senior engineers spend their time on the judgment call: does this fix match how the system is actually built?

AI can speed up the repetitive parts of security work. Accountability for the result stays with the team, and with the people who sign off on it.

Read more: GitHub Changelog, agentic autofix for code scanning alerts (https://github.blog/changelog/2026-10-15-agentic-autofix-for-code-scanning-alerts-general-availability)

Agentic autofix for code scanning alerts is now generally available. It remediates alerts by working across your codebase the way a developer would: it explores relevant files, proposes a fix,…

10/19/2026

Shift Lab still runs the disciplines it always has: product strategy, design, technology, project management and review. What changed is the repetitive work inside each stage. It now runs through a versioned pipeline of AI agent skills, and every project pulls those skills live from one served system rather than keeping its own static copy.

The point is consistency and focus. Routine work moves faster and the same way every time, so senior people spend their hours on judgment, architecture and the client relationship. Deliberate human checkpoints, from spec sign-off to production sign-off, still gate the work.

We instrument the whole pipeline. Across the last 30 days, the serving endpoint handled 2,219 calls with no endpoint failures. Few agencies can put a number on the reliability of their own process.

10/16/2026

shadcn/ui moved its cn helper into a standalone package this September. Components now import it from there, call sites don't change, and a migrate command cleans up the old clsx and tailwind-merge dependencies.

It's a small update with a useful lesson for design systems built on copy-in components. Ownership comes with a maintenance schedule. Every change upstream is a decision about whether and when to bring it into your codebase.

For the design systems we build with Tailwind CSS, Storybook, and component libraries, that decision is a documented part of the handoff. Each client knows which pieces they own, which come from upstream, and how upgrades get reviewed.

Design intent stays intact because someone is accountable for it, long after launch.

Read more: shadcn/ui Changelog, cn (https://ui.shadcn.com/docs/changelog/2026-09-cn)

Components now import cn from the cn package. One dependency instead of two, and no local helper to maintain.

10/15/2026

Between September 7 and October 11, Shift Lab's team completed 959 story points of work across active engagements, org-wide. Weekly totals ranged from 60 to 270, and we report the range as it is.

Shift Lab runs the same disciplines it always has: product strategy, design, technology, project management, and review. The repetitive steps in each now run through a structured, versioned pipeline of AI agent skills, served live from one system so every project follows the current process.

Human checkpoints still gate the work, calibrated to each project. That leaves senior people free for architecture, design intent, and the client relationship.

We can cite a number because we instrument the hell out of our pipeline.