So where are all the AI apps?
TL;DR Highlight
No visible inflection in PyPI package creation after ChatGPT launch — structural reasons why AI productivity gains do not translate into more public software
Who Should Read
Developers questioning AI tool ROI; engineering leaders evaluating AI adoption impact
Core Mechanics
- No clear inflection in new PyPI packages post-ChatGPT — spam/malware spikes excluded
- Post-ChatGPT packages update more frequently in their first year (6→13 releases/year)
- Key reasons: AI-built apps stay internal; first 90% easier but last 10% harder (large codebase, low familiarity)
Evidence
- Answer.AI analyzed total PyPI package count and update frequency by cohort across the top 15,000 most-downloaded packages
- HN comments: iOS App Store new submissions up 24% — metric-dependent picture
How to Apply
- Measure AI productivity gains via internal deployments and automation increases rather than public artifacts like package counts
- After adopting AI tools, invest separately in maintaining debugging competency — technical debt compounds when code is shipped without understanding
Terminology
Related Papers
Can LLMs model real-world systems in TLA+?
LLM이 TLA+ 명세를 작성할 때 문법은 잘 통과하지만 실제 시스템과의 동작 일치도(conformance)는 46% 수준에 그친다는 걸 체계적으로 검증한 벤치마크 연구로, AI 기반 형식 검증의 현실적 한계를 보여준다.
Natural Language Autoencoders: Turning Claude's Thoughts into Text
Anthropic이 LLM 내부의 숫자 벡터(활성화값)를 직접 읽을 수 있는 자연어로 변환하는 NLA 기법을 공개했다. AI가 실제로 무슨 생각을 하는지 해석하는 interpretability 연구의 새로운 진전이다.
ProgramBench: Can language models rebuild programs from scratch?
LLM이 FFmpeg, SQLite, PHP 인터프리터 같은 실제 소프트웨어를 문서만 보고 처음부터 재구현할 수 있는지 측정하는 새 벤치마크로, 최고 모델도 전체 태스크의 3%만 95% 이상 통과하는 수준에 그쳤다.
MOSAIC-Bench: Measuring Compositional Vulnerability Induction in Coding Agents
티켓 3장으로 쪼개면 Claude/GPT도 보안 취약점 코드를 53~86% 확률로 그냥 짜준다.
Refusal in Language Models Is Mediated by a Single Direction
Open-source chat models encode safety as a single vector direction, and removing it disables safety fine-tuning.
Show HN: A new benchmark for testing LLMs for deterministic outputs
Structured Output Benchmark assesses LLM JSON handling across seven metrics, revealing performance beyond schema compliance.