- Generative AI - Short & Sweet
- Posts
- Gemini 2.5 Pro One-Shots a CRM?! + Manus Reality Check (Win Access)
Gemini 2.5 Pro One-Shots a CRM?! + Manus Reality Check (Win Access)
2 exclusive access codes available
Martin Musiol
April 02, 2025 • Estimated Reading Time: 10 minutes
In partnership with
Hey there,
Google has built a monster LLM. Each time I use it I am delighted, and occasionally astouned. So, I am sharing the specs and use cases I found eye-opening.
Also, I stress-tested Manus (I got access after 3 weeks).
🚨 I have 2 access codes for Manus. Want one? Reply to this email & tell me what content you want more of in this newsletter. I'll pick two winners based on the feedback. Fair? Reply! 📧
Your job called—it wants better business news
Welcome to Morning Brew—the world’s most engaging business newsletter. Seriously, we mean it.
Morning Brew’s daily email keeps professionals informed on the business news that matters, but with a twist—think jokes, pop culture, quick writeups, and anything that makes traditionally dull news actually enjoyable.
It’s 100% free—so why not give it a shot? And if you decide you’d rather stick with dry, long-winded business news, you can always unsubscribe.
(✨ If you don’t want ads like these, Premium is the solution. It is like you are buying me a Starbucks Iced Honey Apple Almondmilk Flat White a month.)
Gemini 2.5 Pro: Google Drops a Bomb
Google's Gemini 2.5 Pro is here, launched in late March 2025, and it's dominating the LMSYS Leaderboard. Having 34 points more than the second place (which is GPT-4o funnily enough above GPT-4.5) is impressive, as it typically get tight in on the top.
This model isn't playing around - the key specs:
A 1M token context window, with plans to expand for 2M. It can ingest entire codebases. Other models are stuck around 200k.
True multimodal input: text, audio, images, video.
"Simulated reasoning": This is its reasoning capabilities. Self-fact-checks for improved accuracy.
It of course scores relatively high if not best on most benchmarks (GPQA: 84.0%, AIME 2025: 86.7%, MRCR 1M: 83.1%). But at this point, it doesn’t matter anymore. The real flex is its applications.
One-shot → office simulation game:
One-shot → Phsics simulation:
One-shot → Health CRM:
Example you can reuse: "Develop a fully functional customer relationship management (CRM) web application tailored for a mid-sized healthcare provider. The application should streamline patient communication, appointment scheduling, record management, and billing processes. As a team of experienced full-stack web developers specializing in healthcare technology solutions, you will design and build this application. The application should feature a user-friendly interface with distinct dashboards for administrators, doctors, and staff. Include functionalities such as patient profiles with medical history, a messaging system, insurance verification, and report generation on key performance indicators like patient acquisition and appointment adherence. The front-end should be built using React. Maintain a professional, solution-oriented, and technically precise tone throughout the development process."
Key advantage is that it’s free.
Why this matters: The one-shot potential is insane. This hints at AIs that could one-shot complex programs on demand. Software-as-a-Service (SaaS) might turn into Service-as-a-Software. 🤯
Manus.im Test Drive: The Chinese AI Agent - Hits & Misses
Finally got my hands on Manus.im, the Chinese general AI agent I mentioned before (newsletter episode). Ran it through code refining, game building, webpage creation from YouTube.
The big plus? It's transparent about its process and suggests 'rules' it learns, which you can edit/approve. Smart implementation.
The reality check:
Webpage from YouTube: Nailed it.
Code refining / Game building: Fell short. It either didn’t finish, or produced errors it couldn’t fix upon request.
Hosting: Inconsistent. Sometimes on its own space, sometimes nowhere. Forget asking for Heroku – didn't happen.
![Left Panel (Notion-style note interface titled “Bonbon Stories webpage”) A user named manus documents issues related to deploying a website for “Bonbon Stories.” They mention creating a ZIP file as an interim solution and executing a shell command (cd /home/ubuntu/bonbon_website && zip -r bonbon_website.zip .) to generate the archive. The message explains the ZIP lets users access the site locally. The note includes bullet points describing the website’s aesthetic features (e.g., playful design with golden yellow, matcha green, and rose pink colors). Right Panel (Markdown checklist in editor) A file todo.md shows a checklist with two sections: • Channel Analysis: Tasks like analyzing the YouTube channel, video content, aesthetics, and creating a report. All tasks are marked with red [x], indicating completion. • Web Page Creation: Tasks like designing the webpage, creating HTML/CSS/JS structure, testing functionality, and deploying. All items are also marked completed.](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/7f6c77ad-93f9-424d-a962-9ff36b69eb79/Screenshot_2025-04-02_at_17.11.40.png?t=1743606721)
Manus in development.
The webpage it coded from a Yt-channel only.
Still, it feels leagues ahead of OpenAI's Operator and other general agents in raw capability, despite the current execution flaws on some tasks. I bet updates fix the rough edges soon.
🚨 Got 2 access codes for Manus.im. Want one? Reply to this email & tell me what content you want more of in this newsletter. I'll pick two winners based on the feedback. Fair? 👍
I hope you enjoyed it.
Happy weekend!
Martin 🙇
I recommend:
Beehiiv if you write newsletters.
Superhuman if you write a lot of emails.
Cursor if you code a lot.
Bolt.new for full-stack development.
Follow me on X.com.
AI for your org: We build custom AI solutions for half the market price and time (building with AI Agents). Contact us to know more.
