Chain of thought

Open AI just released a new model (o1, previously codenamed strawberry) with chain-of-thought reasoning. In practice this means that it plans out how it will respond before responding. The resulting slowdown in response is offset by a significant improvement in maths, programming and science (although much better in physics than chemistry and biology). Interestingly it has reduced performance in English, with ChatGPT-4o outperforming it.

The big difference is the shift from pre-training (make the model as complex as possible so each query is easy) to inference (do more computations on each query). You’ve probably heard about the expense of training these new models and how they have ever-increasing billions of parameters. Researchers are now shifting the other way, believing smaller models might be sufficient if you add more compute power per query.

Inference moves us along the chain towards AGI. OpenAI has five levels for AI capabilities:

1. Chatbots - conversational language

2. Reasoners - human level problem solving (we’re here now)

3. Agents - systems that can take actions (some interesting new products coming out - will see a lot more of these over the coming months)

4. Innovators - AI that can aid in invention

5. Organisations - AI that can do the work of an organisation

With this new model, it unlocks agent capabilities. I think we’ll start to see an explosion of agents for specific roles over the coming months with products that can work across multiple systems and activities. The first versions will likely be limited but in a year or two I think agents will be widespread.

Interesting times!

This Week’s Updates

Enabling the Team

Founder Mode by Paul Graham
After recent comments from Airbnb CEO about how he is getting much more hands on, Paul writes about how he believes founders should be in the weeds.

Coaching Founder Mode by Marty Cagan
Marty expands on Pauls essay to highlight how founder mode is different to micro-managing.

Product Direction

6 Ways to Bring Strategy into Your Work Every Day by David Lancefield
Day to day work can be overwhelming. David shares that by mastering small decisions you can have big impacts.

Begin With The End In Mind by Felip Castro
Start by defining a clear outcome, why it is important (context) and any constraints and guardrails. Then work backwards to discover how to reach it.

Continuous Research

Layers of Product Discovery by Jim Morris
People like what they are comfortable with, but if the PM takes the path of least resistance there will be blind spots.

Deep Dive into In-depth Interviews for User Experience Research by Uwem Usa
Uwem shares a deep dive on how to extract invaluable insights to improve user experiences, from research questions to analysis.

Continuous Design

How to Defend your Design Process by Vitaly Friedman
Vitaly shares how to address unrealistic expectations and foster a shared understanding with stakeholders.

Things To Think About For Being A Service Design Lead by Courtney Maya George
What is involved in the jump from Senior Designer to Design Lead? Catherine shares the details.

Continuous Delivery

AI Search: The Bitter-er Lesson by Aidan McLaughlin
In 2019 Chess Zero was released that quickly became the world's best. But Stockfish, the previous non-AI model quickly regained the crown using AI with search (aka inference).

Look out, kids: PHP is the new JavaScript by Dave Kiss
Software frameworks today merge server and client code. But that complexity comes at a cost. What if we go back to the old ways?

UXDX EMEA:
Last ticket price increase

Save €150 on your ticket. Price increase next week. If you want to join us for the last UXDX in Dublin then don’t miss this opportunity.

FREE COMMUNITY EVENTS 

Check out more upcoming community events 

Video of the Week:
Transforming the New York Times: Empowering Evolution through UX

Learn how to balance user, business, and editorial needs, how design can help simplify complex information, and practical strategies for incorporating new features while maintaining a seamless experience. With real-world examples, such as the integration of games into the news app, Libby and Kristen offers actionable insights for anyone looking to enhance product design in fast-paced, content-rich environments.. Check it out now👇👇

The Results of Last Week’s Poll

The question: Where to focus next?

Unlike some polls we’ve done where there is a bit of ambiguity, this one is clear. I’ll run a series on product strategy over the next few weeks. Stay tuned.