This past week, I deliberately absorbed the news rather than repeating it. The release of the DeepSeek model and app has taken the tech world by storm. The financial markets reacted immediately and strongly to the news that the Chinese open-source model outperformed OpenAI’s GPT-4o model and that leading Western AI labs are overpaying for resources while falling behind. However, several additional details have emerged since.
It also shows even more that Large Language Models are a basic technology that is quickly being commoditized. But, how much should you worry about these first-order concerns that we hear about in the news? It depends on your role and objectives. Second-order questions are much harder to answer. They will also remain truly independent of the model that is currently at the top of the leaderboard. But what are these questions?
Accelerating the Pace of AI Hype Through Weekly News Cycles
By the way, since the hype around DeepSeek started, Alibaba has released an improved model (Qwen-2.5), and OpenAI has just shipped another agent (Deep Research) to compete with the AI-enabled research tool Perplexity (which uses DeepSeek).
But unless your core job responsibility is developing LLMs for a major AI lab, you can safely drown out 3/4 of the noise (uhm, news). January alone has had at least three major news cycles, each lasting about a week:
OpenAI: We are working on AGI and know how to achieve it1 (we also have our own definition of AGI2)
Sam Altman: Agents are everywhere, and we will see the first ones entering the workplace in 20253; Microsoft: by the way, they’ll replace traditional SaaS apps4
U.S. government (plus OpenAI, Oracle, and Softbank): The U.S. is investing $500B in data centers and infrastructure to build out AI5
Financial Analysts: Chinese AI model DeepSeek is better, cheaper, and open source6 (but more details are still forthcoming); OpenAI: DeepSeek models were most likely trained on output generated by OpenAI’s models7 (which is against the usage policy, but tough to enforce abroad)
Keeping up with all these news items and the perspectives surrounding them quickly becomes exhausting—and this doesn't even include the plethora of research papers published every day, or the guidelines of the EU AI Act (e.g., mandatory AI literacy training for employees) that went into effect this week. It just gets harder and harder to prioritize what truly matters.
Thank you for reading The AI MEMO. Feel free to share it with your network.
ONLINE COURSE — Agentic AI: Challenges and Opportunities for Leadership
Prepare for the challenges and opportunities of agentic AI as you lead AI transformation in your organization. Maximize the benefits of these technologies while minimizing disruption.
Asking Second-Order Questions to Realize Business Impact
If you are currently in an AI leadership role or advise leaders on using AI in their business, the pace of innovation is a gift and a curse. Core components such as LLMs are getting cheaper, enabling more innovation at an even faster pace and a higher volume. Similarly, your leadership team and board of directors will demand you translate these scale economies into a tangible return for your company.
But you simply can’t afford to act based on the news of the day. Additional news breaks within a few days. (In the case of DeepSeek, look for reports on operational vs. usually breaks training costs, data privacy/ legal terms, guardrails, and benchmarks.)
Your business’s technology strategy defines the layers that your technology stack is comprised of and which vendors you use for each layer. Once it is in place, your team might evaluate individual vendors and models, but you should spend most of your time on the next level of questions centered around how to create value from AI.
The following aspects are relevant in a business context:
Infrastructure
Models are eventually getting smaller and less resource-intensive. This makes them more affordable to host and run, enabling widespread innovation across industries and companies.
Should you expand from a single- to a multi-vendor strategy, if LLMs are being commoditized further?
How should your technology strategy evolve if hosting open-source models in your cloud is getting cheaper?
Roadmap
Which AI project and products to pursue is an aspect you directly influence as an AI leader. Put high-level news themes into the context of your business and industry to anticipate their impact.
Which AI scenarios will deliver the highest return/ value for your business?
What can you deliver within a finite budget? What else can you deliver if the LLM cost drops?
Can you repurpose your budget to scale projects faster via external resources when the LLM cost drops?
Competition
When prices drop, demand will surge even more than previously anticipated (aka Jevons paradox in economics), which will lead to to additional follow-on effects.
If everyone has access to cheaper models, what will they build? Which part of your business will likely be disrupted?
What new market entrants will be formed that do not even exist today?
You quickly realize that the minutiae of the news cycle become line items in a higher-level strategy. Focusing on second-order questions leads to answering questions of higher value and impact.
Conclusion
While it‘s important to be aware of general developments in AI, focusing on the minutiae of every news cycle distracts from your actual AI leadership role, which is to make AI useful in the context of your business.
First-order concerns about dominant AI models, players, and countries should be replaced by second-order concerns, such as identifying and prioritizing promising AI capabilities that realize tangible results, and third-order concerns, such as the impact on individuals, industries, and societies. The higher the order of the concern, the less important the minutiae of this week’s news cycle become.
Good luck staying focused while making it through this week’s news cycle. We all need it.
Join my bi-weekly live stream and podcast for leaders and hands-on practitioners. Each episode features a different guest who shares their AI journey and actionable insights. Learn from your peers how you can lead artificial intelligence, agentic AI, generative AI & automation in business with confidence.
