Dera News
Back to Intelligence
dera monthly editionJanuary 2026 / Vol.12〜16

The month when letting AI do the work stopped sounding like a joke.

January was not just a story about smarter models. Anthropic shook the assumptions behind SaaS valuations, OpenClaw put AI agents on individual desks, and Google began shipping AI into the browser by default. Delegation stopped being an experiment and became an operational question.

Issue file

Date

January 2026

Editor's Note

January was not just a story about smarter models. Anthropic shook the assumptions behind SaaS valuations, OpenClaw put AI agents on individual desks, and Google began shipping AI into the browser by default. Delegation stopped being an experiment and became an operational question.

The urgent task is no longer deciding whether to use AI. It is deciding which work to delegate to AI and which work humans must continue to own.

What changed this month

Eleven plugins erased roughly $285B in market value

Before: From November through December, the story centered on model gains and enterprise integrations. AI was still mostly framed as an assistant layered into existing software.

Now: In January, AI stopped merely augmenting software and started eating into software's role directly. Markets priced that in almost immediately.

SaaS market value erased $285B

A capable but not fully trustworthy junior arrived for a few dollars a day

Before: Up to December, autonomous agents were mostly discussed as enterprise integrations or future potential, not as tools ordinary people could run directly.

Now: In January, an individual-developer tool exploded into the mainstream, making personal AI agents testable by anyone while proving that cost controls and sandboxing are mandatory.

GitHub stars in 48 hours 140K+

AI quietly entered the browsers of 3 billion people

Before: Back in November, the competitive arena still looked like benchmark scores and raw model capability.

Now: By January, the real contest was shifting toward who could embed AI into the browser, operating system, and day-to-day work environment by default.

Chrome users 3B+

The issue

Escalation01Supports dera's view

Eleven plugins erased roughly $285B in market value

Deep Dive

On January 12, Anthropic introduced Claude Cowork. It could handle file operations, draft documents, and execute workflows directly on a PC. This was not another chatbot. It was positioned as an AI co-worker that acts on the screen.

Then on January 30, Anthropic released 11 open-source plugins for contract review, NDA handling, sales research, financial modeling, and other tasks.

The market reaction was immediate. About $285B in SaaS and IT market value disappeared. Jefferies traders called it the SaaSpocalypse, and Bloomberg amplified the shock.

I have zero programming background, but I used Cowork to control Claude Code and finished in one night an app I had been stuck on for four months.

Reddit user

Counterpoint

Bank of America stayed measured: AI may disrupt task-level knowledge work, but that does not mean it can replace SaaS platforms that manage critical business operations.

Escalation02Supports dera's view

A capable but not fully trustworthy junior arrived for a few dollars a day

Debate

A hobby project from an Austrian solo developer became one of January's loudest stories in AI. OpenClaw crossed 140,000 GitHub stars within 48 hours, one of the fastest ramps ever.

Its core value was simple: let AI operate your own computer. Email handling, browser workflows, scheduled tasks, inventory checks - the agent could act remotely through Telegram.

But reality arrived fast. Daily use could cost the equivalent of $15 to $50. Security looked worse: about 40,000 exposed servers and 341 malicious plugins were identified.

I save one to two hours a day just on email triage and draft replies. The real shift is not one feature - it is having something that works 24/7, remembers context, and can act while I am away.

Hacker News user

Counterpoint

One security roundup went so far as to call it the biggest security disaster in AI history.

Reframed03Revises dera's view

AI quietly entered the browsers of 3 billion people

Contrarian

At the end of January, Google introduced Chrome Auto Browse, a deep Gemini 3 integration that lets AI operate the browser autonomously.

Chrome Auto Browse plugs into Google Search, Workspace, and Android. Google's massive search and productivity footprint becomes the context layer for the browser agent.

That means Google is no longer fighting only on model performance. It is fighting on switching costs. If a company already lives in Gmail and Drive, Auto Browse starts looking like the default they never have to choose.

You are responsible for Gemini's actions during the task.

WIRED

Counterpoint

Early Reddit feedback was harsh: too buggy, opening windows users never asked for, and unable to handle follow-up questions in the middle of a task.

Decision Board

Questions and execution guidance

This is the execution section: finalize what to test now and what to govern before rollout.

This Month's Questions

  1. 1Who decides which work your company can delegate to AI and which work must remain human-owned?
  2. 2Each time you accept convenience, are you conscious of what you are giving up in return?

Editorial View

  1. 1The urgent task is no longer deciding whether to use AI.
  2. 2It is deciding which work to delegate to AI and which work humans must continue to own.

Continuity

How the monthly arc moved

This section shows where the monthly arc stood as of January 2026. It is the issue's endpoint at the time, not a live status view.

November 2025

Model war

Model War

December 2025

Reality check

Reality Check

January 2026

Delegation

Eleven plugins erased roughly $285B in market value / A capable but not fully trustworthy junio…

Nearly selected

What almost made the issue

This issue closed tightly around the final three stories.

Open questions

Who decides which work your company can delegate to AI and which work must remain human-owned?

Each time you accept convenience, are you conscious of what you are giving up in return?

Method

This edition was compiled by reconciling dera editorial memory, dera reporting from the month, and outside community research.

Colophon

This edition was compiled by reconciling dera editorial memory, dera reporting from the month, and outside community research.

Source notes

Monthly note

📩 This report is published monthly. Subscribe to dera news (free)

💼 Want help implementing AI? Talk to dera

Week by week

Innovation weekOptimistic

Week 5

Google is transforming the web browser from a passive viewing window into an active 'agentic workspace' with Chrome's new Auto Browse capabilities, while the enterprise sector decisively shifts toward a 'multi-model' strategy—exemplified by ServiceNow's pivot to Anthropic—to avoid vendor lock-in.

Articles
8
Companies
12
Topics
30

As Chrome becomes an agent and enterprises curate model portfolios, are we witnessing the transition from 'using' software to 'managing' intelligence?

Expansion weekOptimistic

Week 4

The AI battlefield is shifting from 'model superiority' to 'interface dominance,' as OpenAI prepares to enter hardware and Apple integrates intelligence directly into the OS, signaling a future where AI is no longer a destination but an ambient layer of reality.

Articles
9
Companies
12
Topics
35

As OpenAI builds hardware and Apple integrates intelligence, are we witnessing the death of the 'app' as the primary unit of software?

Structural Integration weekOptimistic

Week 3

The AI industry is simultaneously scaling up and drilling down: massive infrastructure bets like OpenAI's chip deal are securing the 'macro' future, while Google and Apple are racing to conquer the 'micro' context of personal data and commerce.

Articles
8
Companies
16
Topics
30

When intelligence becomes industrial infrastructure, does the true value shift from the model itself to the personal context it consumes?

Milestone weekOptimistic

Week 2

The AI industry has crossed a critical threshold from 'chatting' to 'doing,' with Anthropic's new ability to control PCs and Microsoft's transactional 'Buy' button signaling the arrival of true autonomous execution, underpinned by massive new infrastructure investments from SoftBank and OpenAI.

Articles
9
Companies
19
Topics
34

When AI takes the cursor and the credit card, does the human role shift from 'operator' to 'supervisor'?

Monthly synthesis

January marked a profound shift in the AI landscape, moving decisively from conversational interfaces to direct, autonomous execution within existing digital environments and physical hardware. This operationalization of AI agency was underpinned by massive infrastructure investments in energy and compute, alongside a fierce battle among tech giants to own the user's personal context and the very interfaces (OS, hardware, browser) through which AI operates.

AI's evolution from 'chat' to 'direct execution' and agency

The month began with AI gaining direct control over computer interfaces (Anthropic's cursor control, Microsoft's 'Buy' button), demonstrating practical agency. This evolved into major players like OpenAI and Apple aggressively seeking to own the hardware and operating system layers (earbuds, overhauled Siri) for direct user access, culminating in the web browser itself becoming an agentic workspace (Chrome's Auto Browse).

Massive investments in AI infrastructure and data industrialization

The month started with significant capital injections into energy and compute infrastructure (OpenAI/SoftBank energy, xAI capital raise). This progressed to formalizing data supply chains (Wikipedia deals) and industrial-scale compute facilities (OpenAI/Cerebras chip deal), ultimately leading to the vertical integration of energy and compute into 'Compute Conglomerates' (Tesla/xAI).

The strategic battle for user interface and personal context ownership

Initially, the competition centered on who could best integrate AI into personal context (Google Gemini reading photos/emails, Apple/Google Siri/Gemini talks). This rapidly escalated into a strategic pivot towards owning physical hardware (OpenAI earbuds) and operating system layers (Apple's Siri overhaul) for direct, unmediated user interaction, and finally extended to transforming the web browser into an active, agentic interface (Chrome's Auto Browse).