Google Killed Inbox. Its Creators Rebuilt It With AI.

PRODUCTIVITY

In April 2019, Google killed Inbox, the email app a small but devoted group of users genuinely loved. Andrew Lee, a former engineering lead on Gmail who helped build it, left Google, raised $9 million from USV and Lightspeed, and created Shortwave — email rebuilt with AI at the architecture level rather than stapled on top of a 20-year-old codebase.

Powered by Claude models under the hood, Shortwave gives you an assistant that organizes your inbox, writes drafts in your voice through a feature called Ghostwriter, searches your email history with natural language, and lets you create filters by describing them in plain English. Type "archive anything that looks like a marketing newsletter" and it just does it.

But the feature people talk about most is bundling — Shortwave groups similar emails together so you can clear promotions, updates, and receipts in bulk instead of opening each one. On Reddit, users praised it for clearing out promotional clutter without losing anything important. The obvious comparison is Superhuman at $30/month. Shortwave starts at $24/month for the Business plan, which closes that gap — though Reddit users who tried both say Superhuman still wins on design polish and raw keyboard speed.

The bigger challenge is Gmail itself. Google is rolling out free AI writing, summaries, and search features on top of roughly 1.5 billion users. Shortwave argues its AI runs deeper, with multiple model calls per request versus Google bolting features onto legacy code. That may be true. But asking people to pay $24/month when free AI email keeps improving is a conversation that only gets harder over time.

Two things to know: Shortwave only works with Gmail — no Outlook or iCloud Mail. And the team ships fast, going from zero AI features to 14 in a single year and maintaining that pace since. Whatever you think of the product, the velocity is real.

If you're a Gmail user buried in email and Superhuman's $30/month feels steep, Shortwave at $24 gets you most of the way there with smarter AI organization and a team that ships relentlessly. But if Gmail with Gemini baked in handles 80% of what you need, paying $288 a year for a small startup's email client is a bet on their survival as much as their product. Try the free version first. If bundles and AI filters change how you feel about your inbox, upgrade. If not, Gmail is getting better every month and it doesn't cost you anything.

PRODUCTIVITY

Read AI is an AI meeting assistant that records your calls, transcribes them, and generates summaries with action items across Zoom, Google Meet, and Teams. On paper, it's a solid productivity tool. In practice, it might already be in your meetings without anyone inviting it.

The way Read AI spreads is the real story. One person on your team signs up, connects their calendar, and if auto-join is enabled, the bot starts dropping into calls automatically. After the meeting, other participants get emailed a summary — and if they then authorize the app, it begins joining their meetings too. One sysadmin on Reddit described returning from vacation to find it all over his organization's tenant. The company has acknowledged the friction but says it has rejected requests to make the bot less noticeable, arguing visibility is better than stealth. The problem is that trust is exactly what's missing.

Trustpilot reviewers give Read AI a 1.4 out of 5, with roughly 84% one-star reviews. Multiple reviewers used the word "virus."

For people who actually choose it, the product works fine — transcription is decent, summaries are structured, and AI search across past meetings is genuinely useful. But the legal exposure is hard to ignore. Recording without consent is illegal in at least 12 U.S. states, and if transcription consistently garbles certain speakers, that's a discrimination liability even without intent.

Meanwhile, the market is moving toward bot-free alternatives. Granola just raised $125 million at a $1.5 billion valuation, and a key part of its pitch is that no bot joins your call. Fathom makes the same promise. The trend is clear: people want meeting notes, they just don't want a stranger sitting in the room to get them.

Skip Read AI. The core features work but you can get comparable transcription from Granola or Fathom without a bot joining your calls, without auto-emailing summaries to external participants, and without the privacy mess that has led some IT departments to block it at the network level. If Read AI is already showing up in your meetings uninvited, talk to your IT team about revoking its permissions in your Zoom or Teams admin console. A meeting tool that multiple reviewers compare to malware—and that the majority of its Trustpilot reviewers rate one star—is not a tool you want anywhere near your organization.

PRODUCTIVITY

Every automation tool in 2026 calls itself an AI agent platform now. We looked at five no-code options that claim to let anyone build AI-powered workflows without writing code. Most of them are doing exactly what workflow tools have done for years — just with an AI model plugged into the middle.

Here's what none of these companies want to say out loud: the "AI agent" label is mostly marketing. What you're actually getting is a workflow tool that calls Claude or ChatGPT at certain steps to classify, summarize, or draft something. That's genuinely useful. It just isn't the autonomous agent revolution the landing pages promise. Three of the five we tested are still worth your time:

+ Zapier is still the safest bet for beginners. Over 7,000 app integrations, templates for practically everything, and new AI features that let workflows classify text and draft responses based on context instead of rigid rules. One stress test found it cut lead triage from 25 minutes to under one. Starts at $19.99/month but costs climb fast with volume.

+ Gumloop is the most interesting new player. Visual canvas, drag-and-drop nodes, and you can describe what you want in plain English and it builds the workflow for you. The subflow feature lets you nest mini-workflows inside larger ones, which nobody else really does. Free plan gives you 2,000 credits; Pro runs $37/month.

+ Relay.app has one killer feature: it pauses workflows for human approval before the AI continues. Need someone to review a draft or sign off on a decision before it goes live? That's what Relay was built for. Starts at $38/month, though the integration library is smaller than Zapier's.

Make ($9/month) gives you more visual control over complex logic than Zapier, but the learning curve is steeper and it feels technical if you've never automated anything before. Starting with the simplest tool that solves your bottleneck almost always wins — picking something powerful when nobody internally wants to debug a broken workflow is how automation projects quietly die.

Start with Zapier if you've never automated a workflow. Try Gumloop's free plan if you want something more visual and AI-native. Use Relay.app if your workflows need a human to sign off before AI acts. Skip everything else until you've outgrown one of these three. And whatever platform you pick, pair it with Claude for the AI steps. The platform is the plumbing. The intelligence comes from the model you plug into it.

EVERYTHING ELSE IN AI

+ Anthropic's Mythos AI reportedly breached nearly all NSA classified systems in hours, not weeks — during a red-team test, prompting the first-ever U.S. export ban on a live AI model

+ Google lost two of its most important AI researchers in one week transformer co-inventor Noam Shazeer left for OpenAI, Nobel laureate John Jumper left for Anthropic, and the stock dropped 5%

+ A single Reddit comment can hijack what ChatGPT and Gemini recommend to you researchers found a 13-word post is enough to poison AI deep research results

+ SpaceX signed a $6.3 billion deal to rent out its AI supercomputer to open-source startup Reflection AI at $150 million a month

+ The Supreme Court struck down federal rules requiring AI hiring audits , ruling 6–3 that agencies can't regulate AI without explicit authorization from Congress

+ Disney and Universal jointly sued Midjourney for letting users generate their copyrighted characters

+ Oracle reportedly cut 21,000 jobs to fund its $50 billion AI buildout and employees say they were asked to train their AI replacements first

+ Google DeepMind signed an AI research deal with indie film studio A24 to explore how AI can support filmmakers, backed by a $75 million Alphabet investment

OTHER TOOLS

+ Wispr Flow: *(sponsored)*: The new viral voice-to-text AI for iPhone and Mac that actually understands what you're saying every time

+ Superpower: *(sponsored)*: The health app that tests your blood and uses AI to tell you what's actually going on with your body before your doctor does

+ Runway: Runway's Aleph 2.0 video editing model is now inside Figma Weave — edit a single frame and the AI carries that change through every frame of your video, no reshoots needed

+ Adobe Firefly: Adobe's AI assistant now works across Premiere Pro, Illustrator, and InDesign, letting creators describe what they want in plain English while it handles tedious multi-step tasks like generating brand kits, cutting rough video edits, and building storyboards

+ Grok Build: xAI's coding agent now has a `/goal` command that lets you hand off an entire task — like migrating a codebase — and the agent plans, executes, and checks its own work until the job is done

+ Google AI Studio: Google's Interactions API just hit general availability, giving developers a single endpoint to build with Gemini models and agents, including background execution and managed agents, in just a few lines of code

+ Productboard Spark: The product management platform launched an AI agent that already knows your customers, roadmap, and business context — it can draft specs, summarize feedback, and write stakeholder updates so product teams spend less time on busywork