Slackbot's AI agent upgrade

Welcome back, Superhuman. Slack just shipped its biggest update in years: a completely rebuilt Slackbot that works less like a chatbot and more like a coworker. It knows your apps, reads your files, and pulls context from across your workspace to do useful tasks.

Today: The new-and-improved Slackbot, learn how to create your new profile picture with Gemini, and get the latest prompts and trending social posts.

TODAY IN AI

1. Slack transforms Slackbot into a full AI agent for work: The Salesforce-owned messaging platform rolled out a completely rebuilt Slackbot that can find your information across connected apps, draft emails, and schedule meetings — all without leaving Slack. The AI agent connects to your existing tools, messages, and files to deliver context-aware answers. It’s currently available for Business+ and Enterprise+ customers, with no new downloads required. Watch it in action here. 

2. Signal founder launches Confer, an end-to-end encrypted AI chatbot: The engineer behind Signal Messenger unveiled Confer, an open-source AI assistant designed to make user data completely unreadable to anyone — including hackers, law enforcement, and the platform itself. The service uses trusted execution environments (TEEs) to encrypt all conversations, with decryption keys stored only on users' devices. Learn more here.

3. Apple launches Creator Studio, bundling six pro apps for $12.99/month: The tech giant unveiled a new subscription for its creative apps, including Final Cut Pro, Logic Pro, Pixelmator Pro, Motion, Compressor, and MainStage — plus premium templates and a new "Content Hub" in Keynote, Pages, and Numbers. New AI features accompany the Jan 28 launch: Final Cut Pro gets Visual Search to find moments by describing them, Keynote can generate presentation drafts from text outlines, and Numbers adds Magic Fill for pattern-based table completion. Learn more here.

PRESENTED BY WISPR

The “Agentic Commerce” trend is exploding—here’s how to take advantage

Typing wastes time. Flow makes voice the faster, smarter option everywhere.

Wispr Flow delivers no-edit confidence:

• 4× quicker than typing. Dictate emails, docs, Slack updates, or code.

• AI auto-edits. Filler words gone, grammar fixed, formatting intact.

• Works everywhere. From Slack, Notion, Gmail, and ChatGPT to Cursor, Windsurf, and iPhone.

• Smarter for teams. Snippet library drops in links, templates, or FAQs on cue.

Flow users save an average of 5–7 hours every week — full workdays back every month.

Give your hands a break ➜  start flowing for free today.

FROM THE FRONTIER

Researchers agree — AI is more like an alien organism than a new technology

The black box problem. Hundreds of millions of people use large language models (LLMs) every day, but not even their creators fully understand how they work. A new MIT Tech Review feature explores how researchers are treating LLMs like alien organisms — dissecting them layer by layer to learn what's happening inside.

Understanding LLMs. Researchers are leveraging two techniques to better understand how LLMs work. Mechanistic interpretability (popular at Anthropic) uses secondary models to trace how activations move through an LLM, almost like an MRI for AI. Chain-of-thought monitoring (OpenAI’s preferred method) reads a reasoning model's internal "scratch pad" as it works through problems — like listening to its inner monologue.

What they're discovering is strange. LLMs don't verify facts the way humans do. When you ask Claude, "Are bananas yellow?" and "Is 'bananas are yellow' true?" — both answers are yes, but they come from completely different parts of the model. One part knows the fact; another evaluates claims about facts. There's not necessarily a central source of truth keeping things consistent — which is why chatbots contradict themselves so easily.

The race for answers. Better understanding LLMs may be one of the most important initiatives in tech right now. Each week, these models become more embedded in everything from hiring to healthcare. But, in reality, we're building a future on top of an intelligence that we don’t fully understand.

THE AI ACADEMY

How to create your new profile picture with Gemini

• Go to Gemini and upload a clear, well-lit photo of yourself.

• Optionally upload a reference image for style, lighting, and mood.

• If you’re using a reference image, paste this prompt:

Sample Prompt: Create a professional corporate headshot using my uploaded photo. Match the overall style, lighting, and aesthetic of the reference image while keeping my facial features, proportions, and identity exactly the same. Dress the subject in [describe your outfit or desired attire]. The final image should look suitable for LinkedIn and professional websites.

• If you don’t have a reference image, use this prompt instead:

Sample Prompt: Generate a clean, professional headshot from my uploaded image while preserving my exact facial features and natural skin detail. Frame the portrait from the chest up with proper headroom. The subject should face the camera directly with a confident, natural expression. Style the subject in a modern professional [attire]. Use a soft neutral studio background in deep charcoal grey (#2A2A2A). Apply gentle, diffused studio lighting with sharp focus on the eyes, natural tones, subtle background blur, and a polished, modern finish.

• Gemini will generate a studio-quality, professional profile photo in seconds. Create one for yourself or share this with someone who needs a new profile photo.

PRESENTED BY SHOPIFY

The “Agentic Commerce” trend is exploding—here’s how to take advantage

Big AI news from Shopify:

• Shopify built Universal Commerce Protocol with Google so any agent can reach every merchant

• Brands can sell in AI Mode in Google Search, Gemini, and Microsoft Copilot

• Agentic Plan opens Shopify Catalog to any platform—no Shopify store required

Learn more about agentic commerce.

IN THE KNOW

What’s trending on socials and headlines today

📝 Claude Cowork: The internet was blown away by how quickly Cowork can organize a messy desktop (10K likes).

➕ Solving Equations: ChatGPT 5.2 Pro just solved a math problem that has stumped humans for decades.

📝 Cowork Review: One user wrote a full review of Cowork that’s racked up over 1M views and 2K bookmarks so far.

🍌 Nano Banana: This Nano Banana prompt helps you create mini 3D characters of yourself or others — with just a selfie.

✈️ Flight Finder: Learn 7 ChatGPT prompts that can help you find cheaper flights (this post has over 10M views).

PRODUCTIVITY

5 New & Trending AI Tools

• 🔍️ Nume*: Nume is the world's first AI CFO that every startup and SME needs. Connects to Xero and QuickBooks in 5 minutes.

• 📄 Pane: Create, analyze, and collaborate on spreadsheets with an AI Agent.

• 🤖 PDF.beauty: Turn Banana Pro image-based PDFs into fully editable PPTX.

• 🔈️Muze: Create and manage your Meta and Google ads with AI.

• 🖼️ Rewarx: Instantly transform raw snapshots into professional 4K commercial assets with AI.

^{* indicates a promoted tool, if any}

PROMPT STATION

The Reverse Brief Prompt

Prompt: I want to achieve [describe the outcome you think you want].
Do not give me strategies, steps, frameworks, or advice yet.
Before proposing any solution, your only task is to deeply understand the problem.

Start by asking me exactly 5 high-quality clarifying questions that help uncover:

Constraints
– limits around time, budget, tools, skills, authority, or external dependencies

Resources
– people, tools, data, budget, audience, systems, or leverage I already have access to

Timeline & urgency
– deadlines, milestones, and whether speed or quality matters more

Success criteria
– how success will be measured, what “done” actually looks like, and what would make this a clear win

The real objective
– identify whether my stated goal is a proxy for a deeper or different outcome (e.g., status, revenue, freedom, validation, risk reduction)

Rules:
Ask the questions one at a time, in logical order
Make each question specific, non-generic, and decision-shaping
Challenge assumptions if something sounds vague or misaligned
Do not suggest solutions, tactics, or examples until all 5 questions are answered

Architectural Specimen

NanoBanana Prompt: A preservation architect's survey board for [BUILDING / STRUCTURE] — [STYLE / ARCHITECT / PERIOD]. Left section: site plan and context, showing the structure in its urban or landscape setting with orientation, shadows, and neighboring relationships mapped. Center section: the building dissected, a cutaway axonometric revealing structural system, material layers, hidden infrastructure, with callouts identifying original versus modified elements. Right section: the building as experienced, photographed from the canonical viewpoint in ideal light, with human figures for scale and life, the architecture inhabited. Visual style transitions from technical survey blue-line through neutral analytical greys to golden-hour warmth. Title block reading "[BUILDING NAME] — [ARCHITECT], [CITY], [YEAR], HERITAGE SURVEY".

Whenever you’re ready to take the next step

• Check out our Top 125 AI Tools

• Get better at prompting with our Top 1,000 Prompts

• Learn how to use AI at work and get certified with our free course

Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 1M+ readers and 2M+ followers on socials. Get in touch.

Until next time — Zain & the Superhuman AI team