In This Issue

• Don’t fall for GenAI deception

• Starting your day more productively

• Time Management in the Age of AI – corporate workshop

Thanks for subscribing to Confirmed Achiever. Please share this with your friends so they can subscribe too

Don’t fall for GenAI deception

If you’ve spent any significant time with ChatGPT, Gemini, Claude, Copilot, or other GenAI models, you’ve undoubtedly encountered “hallucinations” – untrue responses. In my experience, many of the hallucinations happen when ChatGPT can’t determine the difference between generalizable info and specific info about a topic – so it synthesizes something to have a response that sounds plausible.

My colleague, Austin Orth, publisher of Pittsburgh Tech Beat, noticed discrepancies on charts as OpenAI’s CEO, Sam Altman, announced ChatGPT-5, the latest model that they announced last week.

In the live-streamed announcement, Altman showed a chart showing “Deception evals across models” in which ChatGPT-5 was 2.4% more deceptive than a previous model (o3). Turns out the data is incorrect altogether – and Orth wasn’t the only person who caught it.

So I dug deeper into the difference between “coding deception” and “hallucinations” in case it was important for users to understand as part of their AI usage.

According to ChatGPT’s own documentation, Coding deception means the model misrepresents what it did (e.g., claims it ran code or used a tool when it didn’t, or asserts success on an impossible task).

The same source says Hallucination = the model states factually false content (wrong facts/claims).

OpenAI goes on to state in its document that it evaluates deception in special setups (agentic coding with impossible obstacles, broken tools, missing images, etc.), and evaluates hallucinations with factuality benchmarks.

The company has fixed the inaccuracy in its charts; but I can’t help but think that this highly technical differentiation between coding deception and hallucination is one of the reasons that we need to be particularly wary of using GenAI without checks and balances related to common sense and source checking. And is one of the reasons it’s sometimes hard to understand the reliability of what you’re getting.

It reminds me of early on, people said “the Internet” as the source of their data – instead of which website, database or writer was responsible for putting the data where you can access it.

My first suggestion to you: whenever you’re asking your favorite artificial intelligence product to give you information, also ask for sources; and when data changes over time, ask for the time it was posted (not the time that AI got it for you). Your answers will be more reliable.

Here’s OpenAI’s 58 page description of ChatGPT-5

Starting your day more productively

Do you wake up every morning and check your email first thing? You may be destroying your day’s productivity right off the bat.

A more productive way to start your day is to check your daily To-do list BEFORE you check your email.

When you check your email before you remind yourself what your priority activities are during the day, you play havoc with the way your brain prioritizes your day. You are no longer strategically oriented – you become “interrupt driven”.

That is, you start to prioritize what others want you to do instead of what you have listed as your most important activities. The to-dos from your incoming emails rise to the top of mind, and your AAA priorities sink to a less relevant level.

Sure; there are some roles in which reacting to others is of supreme importance – such as when you’re on a customer support team. Even in those roles, you’ll likely have one or two activities that need to be done before jumping into the support calls.

Frankly, making this behavior change is difficult, because you’re probably changing a habit that is highly ingrained in the way you think. But for most people it only takes a minute daily – or actually doesn’t take any time, because you check your to-do list anyway; it’s just a reordering, not an additional task.

Try it, and I think you’ll likely find your weeks less stressful as you get more of your own priorities done -- probably without deadline pressures.

Time Management in the Age of A.I.

A workshop program for teams that helps them create their own customized methods of optimizing their time by combining the proven popular frameworks with customizations and new techniques like AI to get fabulous gains in their success.

Learn how you can have David conduct it at your organization

This workshop grew out of a successful program first conducted for Executive MBAs at a prestigious university. It’s interactive – and from the comments we get from participants, it’s both fun and impactful:

• "Such a great session! Very interesting on all the different styles of time management. I can't wait to start.”

• "Amazing session. Totally loved it.”

• "This session is essential in life.”

• “You hit the nail on the head”

• “The time flew by — so much info. When I noticed the time, the session was almost over”

The last word

Follow Confirmed on LinkedIn for more interesting business/technology stats, info and tips.

• Your colleagues can subscribe here.

• Get help from our team or find a great tool to achieve every day here.

• Unsubscribe or change your profile at the bottom of this letter.

BTW – If you have a team, confirm a conversation here to explore how you can customize a skills-enhancement program that gives each person the topics and levels that are matched to their roles.

David