Writing Association (LTD)
Kainat Fatima
I’d: 457823
Topic : Technology Growth And Efforts
Main Points:
• Growth is driven by feedback loops, not one-off breakthroughs:
Technology improves when each version teaches you something for the next version.
Example: Tesla’s Autopilot. Each car on the road sends anonymized data on edge cases →
engineers update the model → update ships to all cars → more data.
Effort: Setting up the data pipeline, labeling, testing, and deployment loop. Without that
loop, you stall.
• Cost curves determine who gets to participate:
Most tech stays niche until the cost to use it drops 10-100x.
• Compute: Training GPT-3 in 2020 cost ∼$4.6M. By 2024, similar capability cost
<$100k due to better algorithms and hardware.
Effort: Algorithmic efficiency research, hardware optimization, supply chain scaling.
Growth happens when outsiders can afford to try.
• Standards and interfaces unlock ecosystems:
Growth accelerates when people stop rebuilding the base layer.
Example: USB, HTTP, SQL, CUDA. Once the interface is stable, thousands of teams build
on top without coordinating.
Effort: Requires early players to agree on “boring” specs instead of competing on
everything. Most companies fail here because they want control.
• Talent density > capital density:
Money can buy hardware. It can’t buy judgment.
Kainat Fatima
I’d: 457823
Topic : Technology Growth And Efforts
Main Points:
• Growth is driven by feedback loops, not one-off breakthroughs:
Technology improves when each version teaches you something for the next version.
Example: Tesla’s Autopilot. Each car on the road sends anonymized data on edge cases →
engineers update the model → update ships to all cars → more data.
Effort: Setting up the data pipeline, labeling, testing, and deployment loop. Without that
loop, you stall.
• Cost curves determine who gets to participate:
Most tech stays niche until the cost to use it drops 10-100x.
• Compute: Training GPT-3 in 2020 cost ∼$4.6M. By 2024, similar capability cost
<$100k due to better algorithms and hardware.
Effort: Algorithmic efficiency research, hardware optimization, supply chain scaling.
Growth happens when outsiders can afford to try.
• Standards and interfaces unlock ecosystems:
Growth accelerates when people stop rebuilding the base layer.
Example: USB, HTTP, SQL, CUDA. Once the interface is stable, thousands of teams build
on top without coordinating.
Effort: Requires early players to agree on “boring” specs instead of competing on
everything. Most companies fail here because they want control.
• Talent density > capital density:
Money can buy hardware. It can’t buy judgment.
