It Still Can't Do My Job: Four Years of Moving Goalposts (2022–2026)
publicznyprofil.github.io
July 3, 2026
8 min read
🔥🔥🔥🔥🔥
44/100
Summary
ChatGPT gained one million users within days of its launch, quickly generating widespread social media content, including screenshots of its failures in compiling code. Observations document public reactions and perceptions regarding the capabilities of AI in job performance since December 2022.
Key Takeaways
- ChatGPT initially struggled with coding tasks, leading to Stack Overflow banning its answers due to low accuracy rates.
- GPT-4 demonstrated improved performance by successfully generating a working version of the Snake game, but skepticism remained regarding its capabilities.
- A startup called Cognition claimed to have developed the first AI software engineer, but a detailed review revealed that the demo was staged and did not reflect real-world capabilities.
- Sundar Pichai reported that over a quarter of new code at Google is generated by AI, but critics argue that this figure includes non-essential code and does not represent true engineering work.
Coverage & discussions
Hacker News · publicznyprofil.github.io
Community discussions
Community Sentiment
Positives
- Some commenters see the advancements in AI as genuinely impressive despite the exaggerated claims surrounding AGI, recognizing the potential for real-world applications.
- The ongoing discussion about AI's capabilities highlights the community's engagement and expectation for future improvements, showing a belief in progress.
- The ability of models to generate more coherent outputs is acknowledged, suggesting a gradual improvement in AI functionality.
Concerns
- Frustration is palpable over the constant moving goalposts of AI promises, with many feeling that claims of imminent AGI and job automation have vastly outpaced reality.
- Skeptics are concerned that fundamental issues like hallucinations remain unresolved, casting doubt on the reliability of AI in critical applications.
- The notion that AI could ever reach AGI is dismissed by several commenters, indicating a significant divide in belief about the future of AI capabilities.
