We just quietly passed the A.I Agent industry inflection point
For EVs it was cost per kWh. For solar it was cost per kWh + efficiency. For AI agents, the inflection point was cost + capability hitting a threshold where any small business or individual can employ them today. GLM5 shattered the OpenAI/Anthropic duopoly, and Minimax M2.7 at $20/month sealed it. We are in Apple II territory now — everyone can run a powerful AI agent on a potato.

For electric vehicles it was cost and battery weight [$/kWh +kWh/kg]
For solar panels it was cost and efficiency [$/kWh + kWh/m2]
Reaching these critical industry inflection points allowed an explosion in use of both technologies across the globe, accelerating us towards a better future.

OpenAI and Anthropic’s duopoly on top tier agentic performance was shattered in February by GLM5, with honorable mentions to KimiK2.5 and Qwen3.5.

Minimax M2.7 released with a $20/month subscription plan last week with tokens sufficient for intensive daily use of multiple agents and text to speech generation on every response.
This suddenly catapults us through an inflection point where any small business or individual can employ A.I Agents today.
And this is before Google research released this week the new memory compression algorithm TurboQuant that pushes towards theoretical limits of whats possible:
TurboQuant proved it can quantize the key-value cache to just 3 bits without requiring training or fine-tuning and causing any compromise in model accuracy, all while achieving a faster runtime than the original LLMs (Gemma and Mistral)
We can only expect more efficient (cheaper) data center inference, and more capable local models in the near future.
As little as two weeks ago, we were in the clunky Mainframe or Apple 1 stage. You had to pay through the nose for cloud inference (including GLM5- I hope you used my money to buy something nice for yourself OpenCode).
Your only alternative was to carefully manage short context windows or other limitations to get quantized versions of open source models working locally .
No more, we are in Apple 2 territory at least. Everyone can now run a powerful A.I agent on a potato. Get ready for the wave that is coming.
While we are on Apple, Steve Jobs’ famous allegory was that:
Computers are the bicycle for the mind.
If this is true, then internet was the moped and A.I Agents are the motorbike.
Wherever we are going we are about to get there much faster.