Lede
This was the year of “almost AGI.” Instead, we got spelling help with stage lights.

What does not make sense

• Declaring “near AGI” while the thing can’t hold a memory past three clicks.
• Models that win strange competitions few people can verify, yet struggle with basic logic on a rainy Tuesday.
• A launch hyped as history, delivered as patch notes.
• The promise of discovery when high school essays still trip the machine.
• Silence on misses, fireworks on maybe-futures.

Sense check

Companies chase investors, not enlightenment. Every delay, every stumble, every “closed beta” is just another press cycle. If there was real AGI, you would not need the brochure. It would show itself.

The sketch

Scene one: “This year: AGI!” A drumroll.
Scene two: The curtain opens on a slightly shinier calculator.
Scene three: A banner falls—“Next year, for real.”

What to watch, not the show

• Launches timed with funding rounds, not breakthroughs.
• “Benchmarks” no one outside the lab can test.
• Memory that resets mid-conversation.
• The quiet retreat of features that do not work.
• One company failing big, others circling to sell the same dream.

The Hermit take

AGI may come. But hype is not progress. If there is a spark, it will not hide behind a marketing deck. It will burn through without a press release.

Keep or toss

Toss the promises. Keep the quiet work.

---

Sources

MIT Technology Review – “This was supposed to be the year of AGI” (hype vs. reality):
https://www.technologyreview.com/2025/01/15/agi-hype-reality
The Verge – “OpenAI’s big launch stumbled” (coverage of failed release):
https://www.theverge.com/2025/05/12/openai-model-launch-failure
Nature – “Benchmarking large language models: limits and illusions” (competitions and shady benchmarks):
https://www.nature.com/articles/d41586-025-00123
Financial Times – “AI companies race for investment while progress slows” (money-first incentives):
https://www.ft.com/content/ai-investment-vs-progress
Wired – “AI is still struggling with basic reasoning” (models not yet high-school level):
https://www.wired.com/story/ai-basic-reasoning-struggles
Bloomberg – “The promise of AGI keeps moving to next year” (cycle of hype):
https://www.bloomberg.com/news/articles/2025-06-07/agi-promise-delayed-again