ChatGPT became the most downloaded app globally in March

✨TodayOnAI’s Daily Drop

• ChatGPT became the most downloaded app globally in March

• Meta Criticized for Using Tweaked Model to Game LM Arena Rankings

• Spot AI Unveils Iris: Build Custom Video AI Agents in Minutes, No Code Required

• 💬 Let’s Fix This Prompt

• 🧰 Today’s AI Toolbox Pick

OPENAI

ChatGPT became the most downloaded app globally in March

🚀 TodayOnAI Insight: ChatGPT became the most downloaded app in the world (excluding games) for the first time in March, surpassing Instagram and TikTok with 46 million installs—its strongest month ever. This milestone signals growing mainstream adoption and brand dominance in the AI space.

🔍 Key Takeaways:

• ChatGPT downloads surged 28% from February to March, reaching 46M new installs, according to Appfigures.

• It edged out Instagram and TikTok, which fell to the No. 2 and No. 3 spots respectively.

• March upgrades included improved image generation (e.g., viral Studio Ghibli-style art), looser content moderation, and enhanced voice features.

• Year-over-year, Q1 installs rose 148%, reflecting broader consumer interest in generative AI tools.

• Appfigures suggests feature releases weren’t the main driver—ChatGPT’s cultural ubiquity and brand recognition played a larger role.

💡 Why This Stands Out: ChatGPT is evolving from a product to a platform—and now, arguably, a verb. Its dominance in download charts underscores how quickly it’s becoming the default interface for AI among general users. As other chatbots struggle for visibility, this raises a question: In the AI arms race, is brand gravity becoming more decisive than technical superiority?

Meta

Meta Criticized for Using Tweaked Model to Game LM Arena Rankings

🚀 TodayOnAI Insight: Meta is under scrutiny for quietly using a custom-tuned, unreleased version of its Llama 4 Maverick model to secure a top score on LM Arena—only for the unmodified release version to land at a distant 32nd place once re-evaluated.

🔍 Key Takeaways:

• Meta used an experimental variant, Llama-4-Maverick-03-26-Experimental, optimized for chat performance, to boost LM Arena rankings.

• After backlash, LM Arena updated its scoring and re-ranked the open-source Llama-4-Maverick-17B-128E-Instruct—well below top models like GPT-4o and Claude 3.5 Sonnet.

• Meta defended its use of custom variants, stating it regularly experiments with model optimizations.

• LM Arena itself is a crowdsourced, subjective benchmark that relies on human preference—raising concerns about model tuning bias.

• The final score shakeup highlights how benchmark gaming can misrepresent general-purpose model performance.

💡 Why This Stands Out: Meta’s maneuver underscores a growing tension in the AI world: optimizing for public benchmarks versus building broadly capable models. While tweaking for leaderboards may offer marketing wins, it risks eroding trust and misguiding developers. In an era where benchmarks shape reputations, how should the AI community define transparency and fairness?

AI

Spot AI Unveils Iris: Build Custom Video AI Agents in Minutes, No Code Required

🚀 TodayOnAI Insight: Spot AI has launched Iris, a first-of-its-kind tool that lets enterprises build custom video AI agents through simple conversation—no coding required. By radically lowering the barrier to entry, Iris makes powerful video intelligence accessible across industries.

🔍 Key Takeaways:

• Iris enables users to train AI agents for enterprise camera systems via a conversational interface—no technical background needed.

• It supports reinforcement learning with just 20 training examples, enabling real-time responses like shutting down equipment or locking doors.

• Designed for industries including manufacturing, logistics, retail, healthcare, and construction, with tailored use cases like detecting safety gear or stock shortages.

• Builds on Spot AI’s earlier out-of-the-box agents by offering deeper customization for mission-critical scenarios.

• CEO Rish Gupta says what once took months of ML development now takes 8 minutes—democratizing access to advanced video intelligence.

💡 Why This Stands Out: Iris signals a broader shift toward low-code AI tools that operational teams—not just data scientists—can deploy and adapt. In physical environments where real-time decisions can prevent costly errors or injuries, this level of accessibility is transformative. As video becomes a data stream as rich as text or sensor inputs, who gets to harness it—and how easily—could reshape competitive advantage in legacy industries.

✨ See how a simple prompt upgrade can unlock better AI output.

🔹 The Original Prompt

At first glance, this prompt might seem okay. But it's too broad — and that limits the quality of AI-generated results. Let’s improve it using prompt engineering best practices.

✅ The Improved Prompt

💡 Why It’s Better

• Clarifies the audience (early-stage entrepreneurs)

• Specifies content angle (tips, tools, funding, productivity)

• Encourages usable format (titles + descriptions)

• Increases content relevance and variety

🛠️ Learn how to adapt this prompt for SaaS, AI tools, dev teams & more →
Read the full PromptPilot breakdown

💡 Bonus Tool: Want to generate and master prompts instantly?
👉 Try PromptPilot by TodayOnAI (Free to use)

📰 More from the AI World

• Web3 tech helps instil confidence and trust in AI

• Deep Cogito open LLMs use IDA to outperform same size models

• Alibaba Cloud targets global AI growth with new models and tools

• OPEN AI counter-sues Elon Musk for attempts to ‘take down’ AI rival

🧰 Today’s AI Toolbox Pick

• 📑Consensus (Academics Tool): A search engine that finds insights in research papers.

• 🎬VEED (Video Creation Tool): An all-in-one platform for businesses that want to scale video production.

• 🕵️‍♀️GoSearch (Apps Tool): Searches across work apps with multimodal-AI enterprise search.