ii.inc (Intelligent Internet)

Company note: ii-inc (in COMPANIES). Founder and CEO: Emad Mostaque, formerly co-founder and CEO of Stability AI until March 2024 (Wikipedia; a Render Network post of 2026 also titles him “Founder & CEO of Intelligent Internet”). Author of The Last Economy.

Open-source-oriented AI company building agents and locally owned “intelligence utilities”.

What ii.inc says it does (ii.inc homepage, 2026-10-01)

  • Thesis: as AI commoditises, “value shifts from access to intelligence toward its deployment and integration”.
  • Champions: locally owned institutions (private companies) that run AI utilities for a community; the site says 10% of equity goes to everyone under 21 at setup (vesting at 21) and 0.5% annually to newborns.
  • Agents for individuals and communities; Integration Engineers embedding AI in organisations; Humanoids and robotics as a service.
  • Founding councils can be joined via the site.

Zenith (corrected: a harness, not a model)

The earlier note called Zenith a “featured project” with an unnamed benchmark. ii.inc’s blog (23 Sep 2026, “Zenith: taking a Flash model into the frontier”) describes Zenith as an open-source harness for long-running agent tasks (GitHub Intelligent-Internet/zenith, 318 stars, updated 2026-09-06), run here with DeepSeek V4.1 Flash (deepseek-family).

  • Benchmark: AutoResearchExam (Bespoke Labs): 24 hours per task on each of 29 ML research problems, graded on an unseen test set.
  • Vendor claim: Zenith + V4.1 Flash passes GPT-5.6 Sol and Kimi K3 and reaches nearly 90% of the leader’s score (the leader being Claude Fable 5.1, per the post) at about 1/90th of the cost per task. Self-reported; not independently checked.
  • Earlier (June 2026) ii.inc blog claim: Zenith took GPT-5.5 from 5th to 1st on the same kind of test, ahead of Claude Fable (vendor claim).

II-Agent (GitHub Intelligent-Internet/ii-agent)

Open-source agent “built for real work, now out of beta”: Apache-2.0, 3.4k stars (repo updated 2026-08-16), bring-your-own-API-key, multiple LLM providers (OpenAI, Anthropic, Google), app/site generation, research, document work, integrations with Gmail, Slack and GitHub, web app at agent.ii.inc. Other repos in the org: ii-researcher (504 stars), CommonGround (multi-agent shared work), Evoke (PostgreSQL BM25 engine), opencode-a2a. II-Agent V1 was released 2026-02-04 (ii.inc/releases). ii.inc’s release page says Zenith topped GAIA at 75.57% (vendor claim).

Other releases (ii.inc/releases, 2026-10-02)

Factory Agent (2026-06-16, generative media canvas), Genii (2026-07-29, personal assistant inside messaging apps), Zenith (2026-09-23). 2025: Common Ground (07-17), II-Commons, II-Medical, II-Researcher, II-Thought.

Recent writing

“The Last Economy, One Year On” (22 Aug 2026), “The Intelligence Age Must Be Built Right” (20 Sep 2026), Common Wealth series (Aug 2026) (ii.inc blog).

Dropped as unverifiable

Governance “Guardian Lattice / Oracle Council”, “Foundation Coin” economics, SAGE engine, II-Medical 8B accuracy, roadmap and adoption numbers from the earlier version; re-add only with sources.

Limits

No funding, headcount, valuation or revenue figures are disclosed by ii.inc, so none are given. Zenith results are vendor-reported (ii.inc blog, 2026-09-23) and were not checked against the AutoResearchExam leaderboard.

ai-agent-platforms-comparison, openai, deepseek-family

Sources (accessed 2026-10-02)