The AI race just shifted gears. Smarter models, easier agents, and a coding tool that's quietly replacing entry-level developers.

Hi! Welcome to the 37th edition of the TomorrowToday newsletter.

We’re here to decode the AI chaos so you don't have to. Think of us as your friendly neighbourhood tech translators - we cut through the chaos, translate the jargon, and spotlight new AI tools that matter for founders, builders, and curious minds.

Buckle up, because the future's moving fast and we're here to make sure you don't get left behind! ⚡

If you enjoyed today’s newsletter, please forward it to a friend & subscribe by following this link.

_{~8 mins read}

/Google /Gemini /AIModels

Google just released Gemini 3.1 Pro, and the tech world immediately started celebrating benchmark scores that mean absolutely nothing to most people. So let's skip that and talk about what actually matters.

When AI companies say their model is "smarter," what they mean is that it can handle harder, messier problems - the kind where a simple answer isn't good enough. Think less "what's the capital of France" and more "here's a complex legal contract, a messy spreadsheet, and three conflicting emails - figure out what's going on and give me a plan." That's where 3.1 Pro genuinely improves on its predecessor. Google says it more than doubled its score on a test that measures how well the model solves completely new logic problems it's never seen before. In practical terms, that's the difference between an assistant that only handles familiar tasks and one that can actually think through novel situations.

For the average person, this shows up in three ways. First, Gemini 3.1 Pro is dramatically better at synthesising complex information - take a pile of data and turn it into something clear and useful. Second, it can generate animated SVGs directly from a text prompt, which sounds nerdy until you realise it means you can describe a visual idea in plain English and get a crisp, scalable animation back instantly. Third, and most importantly, it handles multi-step reasoning tasks without losing the plot halfway through.

The model is rolling out now for Google AI Pro and Ultra subscribers via the Gemini app, NotebookLM, and developer tools like AI Studio. If you're already paying for Google One AI Premium, you get it automatically.

/Manus /Meta /Agents /WhatsApp /Telegram

If there's one theme defining 2026, it's personal AI agents. Not chatbots that answer questions - but AI that actually does things on your behalf. Everyone in tech has been buzzing about OpenClaw (the open-source lobster-emoji agent that went viral earlier this year), but here's the honest truth: OpenClaw is powerful and deeply impressive, and also genuinely complicated to set up. It requires installing software on a server, configuring a language model, navigating a command line, and hoping nothing breaks. It's not for your average person.

Enter Manus Agents, now owned by Meta after a roughly $2 billion acquisition at the end of 2025. This week, Manus launched a feature that brings a full AI agent directly into Telegram - with WhatsApp, Slack, and Discord support coming in the next thirty days. The setup? Scan a QR code. That's genuinely it.

Once connected, your Manus agent remembers your preferences, writing style, and work patterns. From a single Telegram message, you can ask it to create a presentation, build a website, generate images, draft emails, update your calendar, or summarise documents. It connects to your Gmail, Google Calendar, and Notion. It processes voice messages, images, and text. You choose your AI model - Manus 1.6 Max for deep thinking, or Manus 1.6 Lite if you're in a hurry.

Here's the key difference from OpenClaw: Manus is doing for personal agents what WhatsApp did for messaging. It's taking something technically impressive and making it absurdly accessible. The power of a full AI agent, inside the messaging app you already use every day.

A quick note: some users on X reported that Telegram temporarily blocked the integration, so expect a few growing pains as this rolls out. But the direction is unmistakable - personal agents that live inside your existing apps are coming whether the platforms are ready or not.

/Anthropic /ClaudeCode /Developers

Claude Code is Anthropic's coding agent that lives in your terminal and writes, reviews, and ships code autonomously. Think of it as the base layer of AI-assisted development - the tool that developers are building their entire workflows around. This week, Anthropic dropped a significant update that makes it considerably more powerful.

Three new features stand out. The first is server previews: Claude Code can now start a development server, open a live preview of your running app directly in the desktop interface, read console logs, catch errors, and keep iterating - without you lifting a finger between each step. The second is local code review: before you push your code to a full review, Claude leaves inline comments flagging bugs and issues. It's like having a senior developer give you a quick sanity check every time you're about to ship. The third is PR monitoring: open a pull request and Claude watches your CI pipeline in the background. With auto-fix enabled, it attempts to resolve failing checks automatically. With auto-merge, pull requests land as soon as they pass - and meanwhile, you've already moved on to your next task.

There's also a new session mobility feature where your coding sessions move with you - start in the terminal, pick it up on the desktop app, and continue on your phone.

Let's be real about what this means more broadly. The work Claude Code now does autonomously - reviewing code, monitoring pipelines, catching errors, iterating on feedback - is exactly the work that used to fall on junior developers. The industry isn't talking about this loudly enough. If you're an early-career developer, the path forward is learning to direct these tools, not compete with them. And if you're a business owner or team lead, the question worth asking is: what could your senior developers achieve if Claude Code handled all the grunt work?

In this section, we aim to spotlight an incredible AI tool or use case and guide you on how you can try it.

If you've been using Claude Code for writing and basic tasks, you've probably started to feel its limits. What if you could give it specialised knowledge for your specific job? That's exactly what Claude Code plugins do - and there are now over 9,000 of them available.

Think of plugins like app stores for your AI coding agent. Instead of Claude being a generalist, you install a plugin and it suddenly becomes an expert in your domain - whether that's legal contract review, financial analysis, marketing copywriting, or automated code review. Anthropic has released 11 official plugins covering Sales, Legal, Finance, Marketing, Support, Product, Data Analysis, and more. They're all free and open source.

Let's use the Legal plugin as an example, because it's genuinely jaw-dropping for anyone who deals with contracts.

Here's how to try it:

The clever part is that the Legal plugin doesn't just read clauses in isolation. It analyses how they interact with each other - so it picks up that an uncapped indemnity clause might be partially offset by a limitation of liability clause elsewhere, and explains the risk in plain English.

Not a lawyer? Also not a developer? It doesn't matter. This plugin was built for exactly that person. Install it once, use it every time a new contract lands in your inbox.

Pro tip: The same principle applies to the Finance plugin for cash flow analysis, the Marketing plugin for campaign briefs, and the Data Analysis plugin for querying your spreadsheets in natural language. Start with the one most relevant to your work.

In this section, we're spotlighting real businesses using AI to solve actual problems.

Meet the kind of story that makes you rethink what a "team" looks like in 2026.

A content marketing consultant - the kind who juggles five retainer clients, a newsletter, and a growing social presence, all on her own - had a problem that will be familiar to anyone who's run a small service business. The work itself was manageable. The admin around the work was eating her alive. Client emails piling up while she was in a creative session. Campaign metrics sitting unread until Friday. Social posts going up late because she'd forgotten to schedule them.

When OpenClaw went viral in January 2026, she was one of the early adopters. After the initial (admittedly fiddly) setup, she connected the agent to her email, her CRM, her task management tools, and her content libraries. Then she wrote it a brief: monitor incoming emails and draft responses to anything standard, track campaign metrics across all five clients and flag anything that looks unusual, repurpose top-performing content for new channels, and compile a morning briefing every day at 7 AM.

The agent now runs overnight. While she sleeps, it's doing competitor research, scheduling social content, flagging the emails that need her personal attention, and generating a daily briefing that tells her exactly what happened while she was offline and what needs her judgment today. She's reported saving 10 to 15 hours weekly - hours that have gone back into client strategy and business development.

The failure modes are real: the agent once sent a client a standard check-in email that should have been escalated, and she caught it only because she reads the morning briefing carefully. The lesson? Treat it like a capable but junior colleague. Give it clear rules, review its decisions, and don't hand over anything that genuinely requires your expertise. But for the mountain of routine work that was previously eating her week? The lobster handles it.

The takeaway: You don't need a team to run like one. With the right agent setup and clear instructions, a solo operator can reclaim double-digit hours per week on the tasks that don't actually need a human - and redirect that energy where it matters most.

AI is full of jargon, and we’re here to decode it. Each week, we’ll give you a plain-English definition of a buzzy term you’ve probably seen (but never fully understood).

Thanks for reading TomorrowToday! We’d love to hear from you:

➡️ What would you like us to cover next?
➡️ Have a tool or topic we should feature?

We’re building this with (and for) you. 🚀
See you next Tuesday 👋