Research & posts

Working paper treating always-on personal agents as a new operating primitive and comparing Dots with Grok Bot.

Short video tip for first-time users plus a note on opt-in tools that let the agent extend itself.

Evidence log of orchestrating Codex, a dot and Muse together, with failure cases and progress-checking methods.

A German-language benchmark comparing Cloudflare Clef Flash 9B, Jev and a 4B model, with test inputs, reference labels, native outputs and evaluation.

Short proofs for six Graffiti conjectures (198, 219, 252, 254, 584 and 712) written with Opus 5.5 and GPT-6.1 Sol.

Code for a Portuguese article comparing Clef against two generative LLMs on 24 hand-labelled support tickets, using the probability as a certainty threshold for automation.

Tournament where teams of agents built from Claude Sonnet 5.5 and GPT-6.1 Sol each write a territory-capture entry, to compare the models head to head.

Four jobs with three attempts each for Claude Sonnet 5.5 and GPT-6.1 Sol, with briefs, checks and every attempt published for replication.

Dataset and scorecard from 3,527 calls across seven tracks comparing Sol 6.1 with GPT-6 Sol, Astra, Luna and GPT-5.6 Sol, with per-item results.

Nandakishor’s dev.to write-up on building decision models a year before Jev, RLCD training against proper scoring rules, and why Laya is open.

Team review of writing, software operation and visual design versus Anthropic's Fable.