A Journey Through AI and Code05 Dec 2025Elena Daehnhardt |
|
Hello, my Dear Reader,
We are celebrating this blog’s birthday again! Elena’s AI Blog is now four years old—still learning, still growing, and still navigating this fascinating AI landscape together with you.
It has been an incredible year since I wrote Three years of Elena’s AI Blog. The AI world has moved at breathtaking speed, and I have been here, learning alongside you, documenting the journey, and sharing my thoughts on everything from multimodal AI to coding assistants.
What is Elena’s AI Blog?
Like everyone today, I live in an era of rapid AI evolution. It is challenging to understand and live in, even for people with a technical background. However, I love making things easy to understand while learning new technologies as a passion. I created this blog to log what I learn and share my ideas and findings.
Now four years old, this blog continues to connect technology with everyday understanding, reflecting my passion for coding and commitment to making complex concepts accessible.
The Blog Since December 2024
Since my last anniversary post, the year has been particularly rich in content about AI coding assistants, multimodal systems, and the practical applications of AI in everyday development work.
1. AI Coding Assistants & Vibe Coding
This year, I discovered something that changed how I code: Cursor AI. I wrote extensively about my experience with this powerful AI-powered development environment.
In Cursor AI for Python Development, I shared my honest journey after months of testing, explaining how it feels like having a brilliant programming buddy. I also explored the concept of “vibe coding” in Vibe Coding with Cursor AI and shared A Few Thoughts on Cursor 2.0.
We also discussed:
- Comparison: AI Coding Assistants
- Methodology: Vibe Coding
- Challenges: Cursor Made Me Do It (Scope Creep) and SaaS survival
- Future Skills: The New Skill Stack: Coding
2. Multimodal AI & Large Language Models
We explored the fascinating world of multimodal systems that process text, images, audio, and video.
- Overview: In Multimodal AI, I explained fusion techniques and models like CLIP and Sora.
- Year in Review: We reflected on the entire year in AI in 2024.
- DeepSeek: I explored running models locally in DeepSeek with Ollama and addressed security in Is DeepSeek R1 Secure?.
- Claude: A comprehensive guide on How to Use Claude AI and a CLI showdown in Gemini CLI vs Claude CLI.
3. Machine Learning Fundamentals & Optimization
I continued to write about the building blocks of ML.
- Core Concepts: Cross-Validation.
- Fine-Tuning: LoRA Fine-Tuning Wins.
4. Practical Python & Development Tools
Beyond AI, we focused on solid software engineering practices.
- Environment: Virtual Environments in Detail and Brewing with Homebrew.
- Git Mastery: Git Log and Should You Use Rebase?.
- GitHub Workflows: Storing Your Local Project to GitHub and GitHub Gists.
- Tutorials: Building a Todo Flask App and using Apache Licensed Summarizers.
We also explored a no-code Workflow Automation with n8n.
5. Ethics, Philosophy, and Reliability
As AI becomes more powerful, we must ask the hard questions.
- Copyright: Who did AI learn from?.
- Philosophy: Could AI Become a New Religion?.
- Reliability: How CustomGPT Minimises AI Hallucinations.
6. Personal Reflections & Experiments
I shared personal moments and fun experiments:
- AI Reads My Blog: Using AI to analyze my own content.
- My Little Setback: Overcoming challenges.
- An Impossible Task for Generative AI.
- Fantastic Voices with ElevenLabs AI.
The Weekly AI News Series
Starting in August, I began a weekly series to keep pace with the breathtaking speed of the industry. This series allowed me to cover agentic workflows, new model releases, and regulatory changes in real-time. Many of the major themes discussed above—from safety to agents—first appeared here:
- AI Heroes of the Week (August 8)
- This Week in AI (August 15)
- Elena About AI This Week (August 22)
- AI Weekly Wins (August 29)
- AI Weekly (September 5)
- AI Weekly News (September 12)
- AI This Week (September 19)
- AI Breakthroughs This Week (September 26)
- AI Got Rules, Wheels & a Lab Coat (October 3)
- Safety, Agents, and Compute (October 10)
- AI Honesty, Agents, and the Fight for Truth (October 16)
- Quantum Thinking, Light Models, Living Networks (October 24)
- Infrastructure, Small Models, and Multi-Agent Coding (October 31)
- Agents Grow Up, Clouds Get Bigger (November 7)
- Ethics, Code, Chips, and a Petaflop on Your Desk (November 14)
- Ethics, Gravity, and the Future We’re Actually Building (November 21)
- Claude Opus, ChatGPT Shopping, and DeepSeekMath (November 28)
Social Networking and Community
As a one-person blogger, I continue to work on disseminating my posts and maintaining traffic from search engines. The weekly AI news series has been particularly well-received, helping readers stay current with rapid developments.
I continue to republish some posts on Medium, following the strategies I outlined in Avoid SEO Penalties on Medium. The platform continues to be a valuable way to reach new audiences.
I post on Pinterest when possible and update my GitHub repositories when I have new code to share.
What I’ve Learned
This year has been extraordinary for AI development. I have synthesised the core structural shifts below:
Annual AI Synthesis Matrix
| Core Theme | 2024–2025 Evolution | Strategic Implication |
|---|---|---|
| Multimodal AI | Matured from experimental to mainstream native processing of text, images, audio, and video. | Data pipelines must now support heterogeneous inputs natively. |
| Coding Assistants | Shifted from passive auto-complete to autonomous, agentic development partners (e.g., Cursor, Antigravity). | Developers must transition from syntax writing to system orchestration. |
| Agentic Workflows | Moving past “chatting” towards delegating multi-step execution chains. | Architectural design now involves writing explicit contracts between agents. |
| Ethics & Safety | Transitioned from philosophical debates to hard compliance frameworks (WHO, enterprise SOC 2). | “Human-in-the-loop” is no longer optional for critical production deployments. |
| Open-Source & Local AI | DeepSeek and Llama democratised access, proving small models can beat legacy giants. | Local, secure AI deployment is now a viable and cost-effective reality. |
| Model Fine-Tuning | Parameter-efficient methods like LoRA made customisation drastically cheaper and more accessible. | General models are commodities; fine-tuned, domain-specific models are the moat. |
Through it all, I’ve learned that the key to working with AI is understanding its capabilities and limitations, using it as a tool to augment rather than replace human intelligence, and always maintaining critical thinking skills.
Next Year Plans?
Looking ahead to 2026, I have several ideas for evolving the blog:
-
Continue the Weekly Series: The weekly AI news posts have been valuable for staying current. I plan to continue and refine this format.
-
More Hands-On Tutorials: I want to create more step-by-step tutorials that combine AI tools with practical projects, showing readers how to build real applications.
-
Deeper Dives into Emerging Technologies: As new AI capabilities emerge, I’ll explore them in depth, explaining not just what they do, but how they work and why they matter.
-
Community Engagement: I’d love to feature more guest posts and create more opportunities for reader interaction and discussion.
Thank you very much for reading my blog, and write to me anytime if you have any comments or ideas or want to say “Hi!”.
Here’s to another year of learning, exploring, and sharing together! 🎈