🍭 Devs Don't Write Code Now 👨‍💻

Good morning. My screen time report asked if I was okay.
I said no and opened three more tabs.

Let’s dive in 👇

🍭 What’s Cookin’:

• Google hits 75% AI-written code

• OpenAI drops workspace agents

• 120K authors file Anthropic claims

Google’s
👨‍💻 Engineers Are Now Code Reviewers

The Bite:

Google CEO Sundar Pichai disclosed at Cloud Next 2026 that 75% of all new code at Google is now generated by AI and approved by human engineers.

The figure was 50% last fall and just 25% in October 2024.

Pichai said the company is shifting to "truly agentic workflows," with engineers orchestrating autonomous agents across codebases.

A complex code migration completed by agents + engineers finished six times faster than it would have a year ago with engineers alone.

Some Google DeepMind employees have also been permitted to use Anthropic's Claude Code internally.

Snacks:

• 75% of new Google code is AI-generated, up from 25% in October 2024

• A complex code migration ran 6x faster with agents than with engineers alone last year

• Google's first-party models now process 16 billion tokens per minute via API, up from 10 billion last quarter

• Gemini’s macOS app prototype was built in days using Google’s internal platform, Antigravity

• Google's Security Operations Center agents reduced threat mitigation time by more than 90%

• Gemini Enterprise saw 40% growth in paid monthly active users quarter-over-quarter in Q1

Why it Bites:

The number that matters isn't 75%.

It's the trajectory: 25% → 50% → 75% in eighteen months.

Google just told every engineering org what "developer" means now:

Someone who reviews and directs machine output.
Not someone who writes functions.

The six-times-faster migration stat is the part hiring managers will screenshot.
And if agents + engineers can do in weeks what engineers alone needed months for…

The math on headcount gets uncomfortable fast.

The risk nobody said out loud at Cloud Next:

When three-quarters of your codebase is machine-generated, the failure mode isn't bad code, but review fatigue.

"Approved at scale" and "understood at scale" are not the same thing.

So, the same company that automates security triage because humans can't keep up with volume is now asking humans to keep up with volume on code review.

Another quiet detail makes the second-order effect already visible:

Some DeepMind engineers are using Claude Code, not Gemini.

When your own people reach for a competitor's tool, the bottleneck isn't the model, but whether any single model can carry a production codebase alone.

And the answer is no.

The agentic era doesn't have a winner yet. It just has a job description change.

Accio Work
💼 Your Agentic Team for Real Business

Meet Accio Work, the agentic workspace for business owners and solopreneurs.

Our specialized agent team manages operations for you:

• sourcing products,

• negotiating with suppliers,

• managing stores,

• and launching marketing.

All on autopilot.

With verified skills and business tool APIs, Accio Work takes action while you stay in control.

Powered by Alibaba.com’s 1B+ products and global supplier data, it handles product selection and execution seamlessly.

Steal This Prompt
💥 Explode Anything Into A 3D Breakdown

This prompt takes any object and blows it apart into a clean, radial 3D explosion. Every component floating in perfect alignment like an Apple teardown got cinematic.
Hard level prompt.

Use it to:

• Turn products into high-end explainer visuals

• Break down complex objects for presentations or content

• Make your startup look 10x more legit than it is

Workflow:

1. Click this link (Prompt)

2. Paste into your AI model

3. Replace the #s with your object, product, its components, and specs

4. Watch it cook like a teardown video directed by Pixar

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🧠 The Autonomous Stack
A plug-and-play architecture to run fully autonomous Claude agents like they’re a startup ops team on autopilot.

🧑‍💻 Claude Code /ultrareview
Spins up parallel AI reviewers to audit your PR like a squad of sleep-deprived senior engineers.

⚡ MiMo-V2.5 & Pro
Frontier-level AI models that hit hard on performance while barely touching your token budget.

🧩 Pazi
Build AI employees for every role in your company and let them plan, execute, and report like they’re on payroll.

🌍 ora
Real-time, on-device translation that turns your Mac into a UN interpreter. No cloud, no strings attached.

_{Everything Else}
🧠 You Need to Know

⚡ Google Splits TPU 8 Into Two Purpose-Built Chips
→ The eighth-gen TPUs launch as TPU 8t for training (9,600-chip superpods) and TPU 8i for inference (3x more on-chip memory), announced at Cloud Next 2026.

🤖 OpenAI Ships Workspace Agents In ChatGPT
→ Codex-powered agents replace custom GPTs for Business and Enterprise plans, running multi-step workflows in the cloud on schedules or via Slack.

⚖️ 120K Authors File Claims In Anthropic's $1.5B Settlement
→ Claims now cover 91% of the 480,000+ works in the Bartz v. Anthropic piracy class action, with final approval hearing set for April 23.

🔬 Google Launches Deep Research Max For Enterprise
→ Two new autonomous research agents built on Gemini 3.1 Pro, scoring 93.3% on DeepSearchQA.

🧠 Google Says 75% Of New Code Is Now AI-Generated
→ CEO Sundar Pichai said AI writes three-quarters of new code at Google, up from 50% last fall, with engineers reviewing and approving all outputs.