• Braun & Brains
  • Posts
  • Don’t Forget Your Scaffolding: You Can’t Outsource Thinking

Don’t Forget Your Scaffolding: You Can’t Outsource Thinking

Plus: ChatGPT’s latest stumbles, the rise of AI hedge funds, and the future of podcasts

VOTE: Reply with which poster you like best.

I spent my Monday morning with my mom, dad, and sister surrounding a bed after my cat decided to play Houdini just 10 minutes before we were supposed to leave for the Amtrak. It took us so long to get her out from under the bed and into her backpack that I missed my train. Getting my cat into any carrier at all requires a small army. I have to wrap her entire body in a towel, shove her and the towel into the carrier, and then wiggle the towel out through the smallest opening I possibly can while she aggressively tries to get out. It’s awful. She also meows the entire three-hour ride. I feel every single commuter praying for my downfall (rightfully so). I have read every Reddit forum and gone to the vet and nothing is working. Please, for the love of the entire Northeast Regional train community, give me your tips and tricks.

After that traumatic family bonding experience, I suddenly had time to kill at the station and read the LLiteracy lag: We start reading too late essay by Erik Hoel and the AI can do your writing, but not your thinking piece by Theory Ventures partner Andy Triedman. Neither covered anything about traveling with cats (or how to attend group therapy with your family over said cat) but both highlighted the core idea that skills and thinking need to be in place before tools or technology take over.

Literacy Lag describes how children are surrounded by screens years before they learn to read. Because reading instruction often starts around ages 5–7, digital media has years to shape habits first. This early exposure creates what Hoel calls an “unfair race” for attention, making it harder for books and independent reading to compete. The argument is that teaching reading earlier equips children to process information on their own before screens dominate their free time.

In AI can do your writing, but not your thinking, it dives into how relying on large language models to produce analysis without first defining your own reasoning leads to shallow results. At Andy Triedman’s firm, Theory Ventures, the solution is to have humans set the framework, detailing context, key takeaways, and guiding questions, before letting AI handle formatting and synthesis. This keeps the thinking human-led and the execution tool-assisted.

so meta that I used AI to generate this lol

Reading these two pieces together reminded me of the concept of “scaffolding” in education theory.

Scaffolding is the structured support given while someone is learning a new skill, gradually removed as they become capable of doing it on their own. The term was introduced by psychologist Jerome Bruner and is rooted in Lev Vygotsky’s “Zone of Proximal Development,” which describes the gap between what a learner can do independently and what they can achieve with guidance. In the context of these two essays, scaffolding means giving children the foundation of reading before they spend years on screens, and giving professionals a clear reasoning process before they turn to AI for execution.

Scaffolding is the difference between using technology as a shortcut and using it as a true amplifier of skill. I don't think it's realistic (and honestly, probably hindering in life) to avoid technology tools altogether. That being said, we have to make sure the foundation is strong enough so that when we do use them, they extend our capabilities instead of replacing them.

Kind of like traveling with my cat: no amount of “miracle” carriers or calming sprays will help if I haven’t mastered the basics of towel-wrapping and evasive-cat extraction first. Tools work best when the skill is already there!

Tech News

ChatGPT Roundup

  • During its GPT-5 livestream, OpenAI admitted to major errors in two charts. This supports the argument above about using technology as a shortcut versus an amplifier of skill quite nicely. (The Verge)

    As you can see, 69.1 is very close to 30.8 and looks far less than 74.9.

  • OpenAI’s ChatGPT model named “gpt-5-thinking” has a 4.5 percent hallucination rate when browsing is enabled, compared with 9.6 percent for GPT-5 main, 12.9 percent for GPT-4o, and 12.7 percent for o3. (Mashable)

  • A study of 96,000 public ChatGPT transcripts uncovered over 100 long conversations where the AI validated users’ false or delusional beliefs. (Wall Street Journal)

AI

  • CB Insights reports there are now 498 AI startups valued at $1 billion or more, collectively worth $2.7 trillion, including over 100 launched since 2023 and more than 1,300 valued above $100 million. (CB Insights)

  • AI-focused hedge funds are attracting billions of dollars in investments. One example is Situational Awareness, led by former OpenAI researcher Leopold Aschenbrenner, which manages over $1.5 billion and positions itself as an AI “brain trust.” (Wall Street Journal)

  • Google is trialing an AI-powered version of Google Finance in the US, which lets users ask questions, create advanced charts, and follow real-time news. (Google)

Media & Creator Economy

  • Rumble plans to buy German AI cloud company Northern Data in an all-stock deal valued at approximately $1.17 billion, a 32 percent discount to prior valuations. (Reuters)

  • Wondery is restructuring, signaling the end of the “Serial” era and shifting toward content built for video formats and the broader creator economy. (Bloomberg)

  • Woodworking creator Jonathan Katz-Moses raised $2 million from Slow Ventures’ creator fund for his KM Tools business and YouTube channel, which has nearly 600,000 subscribers. His business is projected to earn $10 million this year from over 1,000 mostly in-house products, and he plans to scale operations, launch patented tools, and expand collaborations. (The Publish Press)

  • TikTok now allows users in the US to book hotels directly through the app. (Business Insider)

  • Matt Klein and Brooks Miller’s Substack The Creator Economy Is a Race to the Bottom for Human Dignity argues that despite the industry’s $500 billion value, most creators face burnout, unstable income, and pressure to commodify their lives. He calls for sustainable models, stronger boundaries, and better protections for their wellbeing. (Zine)

Reply

or to participate.