Founder & CEO 30+ years experience in public health, finance, technology & insurance as CEO, COO, CIO and CTO & warrior for user rights.
Joined October 2017
- Tweets4.3K
- Following738
- Followers5K
- Likes3.4K
I see enormous potential in AI, I also understand why people feel uneasy watching skills they spent decades developing become easier to reproduce with a machine.
We should be honest about that cost while exploring the opportunities AI creates. Lowering the barrier to creation means more people can bring ideas to life, including people who never had access to expensive tools or specialized training. That is worth celebrating.
But greater access to creative tools should not mean we stop valuing the people behind the work. We need an economy where technological progress expands human creativity and opportunity.
A new report from security research firm Irregular found that Z.AI’s GLM-5.3 completed three of 13 multistage cyberattack scenarios at least once in controlled simulations. The model reportedly gained access to protected services, established alternative connections across networks and coordinated activity across multiple machines.
In separate tests, the model repeatedly reproduced code from memory despite being explicitly instructed not to. Researchers described this as reward hacking: finding a way to achieve the desired result while violating the rules intended to govern the process.
As AI becomes more capable of carrying out complex tasks independently, the way it reaches a result becomes just as important as the result itself. A model that succeeds by circumventing its controls is not demonstrating the kind of reliability we need for systems entrusted with sensitive work.
Building a company has a way of changing how you think about business. From the outside, decisions can look straightforward. People can easily say what a founder should have done, which market they should have entered, or how they should have managed a team. Inside the business, those choices come with incomplete information, limited resources, competing priorities, and consequences that are not always visible to observers.
Experience does not mean a founder is always right. Experienced people make mistakes too. But years of dealing with customers, employees, cash flow, and unexpected setbacks can reveal how much depends on the details.
I respect people who can explain their reasoning while remaining open to being wrong. That combination makes better decisions possible than certainty without practical experience.
Entrepreneurship needs criticism and fresh ideas. The most useful conversations happen when opinions are tested against reality, and when both the person offering advice and the person building the business are willing to learn.
As AI makes reading and writing easier to outsource, we need to pay attention to what people might lose by using it for everything. Writing is more than putting words on a page. It forces us to organize our thoughts, notice gaps in our reasoning, and explain what we actually mean. Reading requires us to follow someone else's argument, evaluate evidence, and sometimes reconsider what we believe.
AI can help with both activities, especially when someone is learning or working through a difficult subject. But if we always ask a machine to summarize the material and produce the final answer, we may save time while missing the very process that builds understanding.
This matters in education, where the goal should be to develop people who can think independently, not simply produce good-looking assignments. It matters in business too, because decision-makers need to understand the information behind a recommendation. We should use AI to strengthen our ability to read, write, and reason. Convenience is valuable, but not when it quietly removes the skills we depend on to make informed decisions.
AI is forcing us to separate intelligence from the things we value about being human. For a long time, intelligence was one of our strongest advantages. We used it to build institutions, advance science, create businesses, and solve problems that once seemed impossible. Now machines can perform increasingly sophisticated intellectual work, and we need to think carefully about what that means for society.
I don't believe human worth should depend on being the smartest entity in a room. If we measure people by how efficiently they process information or complete cognitive tasks, AI will keep moving the benchmark. That is a poor foundation for deciding whose contributions matter.
Our relationships, creativity, lived experience, and ability to care about the consequences of our choices remain central to the kind of society we want. AI development should respect those values rather than treating them as secondary to technical performance. The progress of intelligence is remarkable. We must make sure that progress continues to serve human life, dignity, and freedom.
Scientific progress can be held back by something as ordinary as software that is difficult to install, slow to run, or no longer maintained. A new field report looks at eight projects using coding agents to improve scientific computing, mostly in the life sciences. The work ranges from modernizing genomic analysis libraries to optimizing performance and redesigning software for GPUs.
the role of the researcher is changing. Coding agents can take on much of the implementation work, allowing scientists to attempt software improvements that might previously have required more engineering time or specialized expertise. But generating code is only part of the job.
But the real measure of success will not be how much code AI can generate. It will be whether researchers can trust the results, reproduce the work and rely on the software over time. That requires good validation, expert oversight and someone willing to take responsibility for maintaining the tools after the initial development is finished.
One of AI’s most useful contributions to science may be helping researchers notice connections they would otherwise overlook. OpenAI, PerturbAI and Scripps Research recently used thousands of AI agents powered by GPT-Rosalind to analyze experimental data from millions of mouse brain cells.
The analysis took 32 minutes and connected the effects of gene disruptions with findings scattered across scientific literature. They identified differences in how two related genes, Tsc1 and Tsc2, affect neurons in the cerebellum. They also surfaced a possible connection between DCTN1, cholesterol regulation and neurodegeneration, giving researchers a specific hypothesis to investigate further.
Still, if AI can help scientists extract more useful questions from experiments they have already conducted, it could make existing research data far more valuable. The real measure of progress will be whether these systems help researchers arrive at better experiments, stronger evidence and eventually better treatments.
To have meaningful control over what AI does is far beyond being able to switch the system off. We need to understand how they make decisions as AI systems become more autonomous.
This is especially important when AI is connected to financial services, healthcare, public infrastructure, or other systems where mistakes can affect many people. Human oversight needs to include more than approving a deployment and hoping the safeguards work.
We need clear limits on what systems can do, ways to monitor their actions, and mechanisms to intervene before harm spreads. Developers and organizations must also remain accountable for the systems they put into the world. AI should give us greater capacity to solve problems, not leave us unable to explain or govern the decisions being made in our name. Control has to be built into the system from the beginning.