AI
How People Are Really Using AI in 2026
- Because AI models mimic human thinking, many people think that they are thinking. This leads to workslop, losing track of intentions, outsourcing thinking, writing less, and developing a false sense of intellectual rigor.
- Emotional support is another common use case, which has its own risks.
- For work, AI is being used for growth, transformation, and efficiencies.
How I Deleted 95% of My Agent Skills and Got Better Results
- The speaker described several ways he tried to be clever (read: overengineered) skills, and discovered less is more.
- He ended up moving away from pi and built a Python state machine to guarantee certain things were happening.
- Enforce, don’t instruct. Prompts start to fade once the context window fills up.
- Guide, don’t prescribe. Tell the model where the landmines are, not how to write code.
- Measure, don’t assume. Find repeatable ways to ensure it’s doing what you want.
- If you’re making your product work for agents, figure out what agents get reliably wrong about your product, measure whether your agent context helps, and thing about agent consumers the way you think about developer consumers.
- There’s lots of FOMO (fear of missing out) around AI, from broligarchs to celebrities.
- An NBC News poll showed that 57% of registered voters believe the risks of AI outweigh the benefits.
- A New York Times poll said 18% of people 14-29 said they felt hopeful about AI.
- The University of Central Florida’s commencement (Gloria Caulfield) speech got boos. It also happened at the MTSU commencement (Scott Borchetta).
- The stereotype is that the younger generation is the first one to adopt new technology. However, unlike the Internet revolution that created more opportunities, the AI revolution is removing opportunities.
- Pope Leo XIV recently wrote a 43K-word encyclical letter Magnifica Humanitas on the concerns of AI.
- A Fox News poll found 8 in 10 think it is extremely or very urgent for the government to address AI regulations.
- A Gallup poll showed 7 in 10 Americans oppose local construction of AI data centers.
- Voters in Utah (and even red states) booed having data centers constructed in their backyards.
- The feeling people experience when they see AI in the wild is disgust. The person using it doesn’t care about us and our humanity.
- A Futurism post stated that tech CEOs have a problem where even their closest allies now see AI as a sign of laziness and dishonesty. Paul Graham (Y Combinator) said, “I have never knowingly finished reading an email signed by a human but written by AI. It feels like being lied to, and who would stand for that?”
- When someone says to you, “That looks like it was made with AI.” Is that a compliment?
Worried About Your AI Coding Bill This Month? Share These 6 Habits With Your Team
- AI is stateless and has no memory. It doesn’t just send the latest message; it sends everything every turn.
- Stop using AI as a command runner (e.g., commit this change, run these commands). Only use AI to run commands if you want it to read the output and act on it (e.g., debug a failing test).
- Don’t make AI build a starter project. There are often existing deterministic scaffolding tools for this (e.g.,
npm create vite). - Limit enabled MCP servers.
- Don’t let chats run forever.
- Do one small thing at a time.
- “Stop sending the tank to buy groceries.” Pick a model and reasoning effort appropriate to the task.