From the RAE dictionary, I’ll keep these two definitions of aging:

1. Existing for a long time or persisting in its state.
2. That existed or took place in the past.

The first one interests us. The second one worries us.

taniwa is a software boutique. We build software for third parties and we also make our own products.

We sell our services with messages like:

Quality-driven
Adapted to the latest technologies
With a team of experienced professionals
We do things with care.

In short, a message not very different from other software companies so we emphasize real project references (actions speak louder than words) and continuous training.

The latter is exhausting but also beautiful.

These last two years we’ve had to shift into another gear to adapt to:

1. The emergence of AI and LLMs (Large Language Models).
2. The flood of tools based on the above.

Our goal is clear: Stay “young” and up-to-date, provide real value to our clients and our products.

How do we stay young?

1. Motivation

We don’t think the emergence of LLMs and AI is a threat. We think it’s an opportunity. There’s a lot to learn that gives business opportunities to us and our clients.

2. Copilot now!

All taniwa members have access to GitHub Copilot. We share how we use it and what we learn among ourselves and at events like primerViernes. Copilot takes away years of aging, of slowness, and for now you “type” faster.

3. Courses

In our case: DataCamp with DataScientist and Machine Learning courses. Access for anyone who wants to Packt subscription. Company roadmap for certifications. Currently with AWS.

4. Proof of concepts

You don’t learn as much from reading as from doing. We do proof of concepts with technologies that interest us and our clients. We incorporate technologies that interest us into our products.

5. Demos

If you explain it, you learn it twice as well.

6. Events

Selection of events and attendance. Inviting our clients and friends to events.

Examples from recent months

Copilot

Already mentioned. We use it daily not only for code but also for documentation. For example, to write this text (5%).

Video Indexer

We’re working on a video indexer for one of our clients leveraging LLM capabilities and “computer vision” to:

• Extract text from videos. Both OCR and transcriptions.
• Extract metadata from videos. For example, element and situation recognition.
• Face recognition.
• Index videos to enable searching by text and metadata. Elasticsearch rocks!

Content Generator

• AI-assisted generation of texts, images, and videos.
• Content creation for social networks, blogs, etc.
• Use of tools and - better yet - frameworks.

LLMs with proprietary content

• Super interesting project for industrial plant design assistance.
• Based on proprietary documentation.
• Long live LlamaIndex.ai.

If you don’t have a plan to improve, it will be easier to let yourself drift and in a few years find yourself outside the value chain with few options to get back in.