Jun 11 2025 ~ 3 min read

AI Breakthroughs 2023-2025: Key Milestones & What’s Next


🔥 Two Years of AI: What’s Happened and Where We’re Going

June 2023 → June 2025

I’ve been closely following developments in the AI space and trying out various tools. Since June 2023 we’ve witnessed an incredible pace of innovation - across LLMs, creative AI, productivity tools, and automation.

Here’s a concise recap of the most impactful events over the past two years and a peek into what’s coming next.


🧠 Timeline: Major Milestones (June 2023 - June 2025)

📆 Mid–2023

  • GPT-4 adoption spikesReasoning and coding saw major improvements.
  • Claude 2 (Anthropic) - A strong safety-focused model enters the space.
  • Copilot & Duet AI - Microsoft and Google roll out AI in Office & Workspace.
  • Meta’s LLaMA 2 - Open-source LLMs become viable for real-world use.
  • Runway Gen-2 & Pika - Text-to-video content gains traction.
  • n8n AI integration - Visual automation with GPT inside custom workflows.
  • Zapier AI Actions (Beta) - Add GPT logic to no-code task flows.
  • MidJourney continues to evolve - Improved prompt control, stylization, and coherence.

📆 Early–2024

  • OpenAI previews Sora - Text-to-video reaches new levels of realism.
  • ChatGPT unified tools - One assistant for coding, browsing, image reading, and memory.
  • DeepSeek-VL and DeepSeek-Coder - China’s open-source models impress with performance.
  • Google’s AI Revival - Launches AI Studio, stream mode, and Gemini 1.5 with long context.
  • Apple AI quietly arrives - On-device models & OpenAI integration.
  • Zapier Central & agent support - Easier multi-step GPT-powered automations.

📆 Mid–2024 to Now

  • GPT-4o launches - Real-time, voice-enabled, multimodal AI for everyone.
  • LLaMA 3 & Mistral 7B - Open models get stronger, faster, and easier to use.
  • AI agents rise - Tools like Devin and AutoGPT start showing autonomy.
  • Text-to-video gets cinematic - Luma Dream Machine, Sora, and Pika advance fast.
  • n8n gets smarter - AI-powered pipelines and agents inside your visual workflows.
  • MidJourney stays relevant - Used by creators and brands alike for storytelling visuals.
  • Google Gemini improves - Streaming responses, Gemini Flash, and AI Studio reshape how we build with LLMs.
  • AI regulation begins - EU AI Act passed, and global frameworks emerge.

🔮 What’s Coming (Late 2025 and Beyond)

🤖 1. AI Agents Go Public

Expect agents that can research, plan, book, and even write code without supervision.

🧩 2. Productivity Becomes AI-Native

AI will handle tasks like summarizing meetings, writing emails, and managing calendars automatically.

🧠 3. Smarter Models

Memory, long context (100k+ tokens), and better reasoning will power GPT-5, Claude 3.5, and Gemini upgrades.

📱 4. On-Device AI

Phones and laptops will run LLMs locally for privacy and speed (no cloud required).

🎨 5. Multimodal Creative Workflows

Seamlessly generate high-quality video, images, and sound from plain text in minutes.### 🎨 6. Voice will become the standard to interact with AI

7. The rise of AI agents

AI agnts will reshape tech standards and systems that aren’t built to work with agents will quickly fall out of relevance.

Headshot of Alvaro Uribe

Hi, I'm Alvaro. I'm a software engineer from Chile 🇨🇱 based in New Zealand 🇳🇿.