Proto-AGI is here

Researchers at evertslabs are very exited after we found out that ChatGpt 5.1 is actually the proto AGI release, meaning that the next release (GPT-6) will be the long awaitedAGI release! Many articles about AGI have already been published at this website. See the archive.

Key Improvements

1. Two variants + automatic routing
   • GPT-5.1 comes in two main modes: Instant and Thinking. OpenAI+2DataCamp+2
   • The system can route your query to the mode best suited to it (so you don’t always need to pick manually). OpenAI+1
   • Instant is tuned for quicker, conversational responses; Thinking is tuned for deeper, more complex reasoning. OpenAI+2DataCamp+2
2. Adaptive reasoning / smarter depth of thought
   • GPT-5.1 adjusts how much “thinking” it does based on the task: fewer tokens + lower latency for simpler tasks, more effort for harder tasks. OpenAI+1
   • Example: For a simple task it may respond in ~2 seconds (vs ~10 seconds for GPT-5 in one test). OpenAI
3. Improved instruction following & response precision
   • Better at obeying constraints like “write exactly 50 words”, “format as a table”, etc. DataCamp+1
   • Also fewer instances of going off-script when given clear constraints.
4. More natural conversational tone & personalization
   • GPT-5.1 is described as “warmer” and more enjoyable to talk to — not just capable but more human-friendly. OpenAI+1
   • New tone/style presets (e.g., “Friendly”, “Professional”, “Candid”, “Quirky”, etc) and sliders for things like warmth, formality, emoji usage. DataCamp+1
5. Coding & tooling improvements
   • Better performance for software-engineering tasks: more reliable code edits, better understanding of instructions, improved coordination of tool use like patching code. OpenAI+1
   • New tools: an apply_patch tool (for making structured code updates) and a shell tool (execute commands) introduced with GPT-5.1. OpenAI
6. Extended prompt caching & efficiency
   • Prompts can now be cached for up to 24 hours, so follow-ups can be faster and cheaper. OpenAI
   • Token-efficiency improvements: fewer wasted tokens/thinking cycles on simple tasks.
7. Better safety, clarity & reliability
   • Model responses are claimed to involve less jargon, be more accessible, and better suited for explaining technical concepts to non-experts. DataCamp
   • The system card addendum notes improved baseline safety metrics and evaluation for things like emotional reliance. OpenAI

Since your work involves coding, AI, data workflows — these improvements have direct re