Open Executive: The AI CEO Repo and Which Claude Models It Uses
By Matt Cullerton, founder of Mavric and Apso
Open Executive is an open source GitHub project from Sente Labs that bills itself as "your company's virtual executive team." A TikTok from @alberta.nyc jokes that you can replace your CEO with it, since it defaults to Claude Sonnet 5 and "most CEO decisions are not opus level anyway." The default model claim was correct when the video posted, but the repo sends the most CEO-like work (strategy, finance, legal and board questions) to Claude Opus.
What is Open Executive?
The repository lives at github.com/SenteLabsAI/OpenExecutive. Its README describes it as "An AI system that acts as your company's virtual executive team" and "a senior advisor with Harvard MBA-level knowledge, customized for your specific business." It is Apache-2.0 licensed, written mainly in Python with a Next.js 15 frontend. On 2026-10-19 it had 5,629 stars and 624 forks.
The video calls it "a new open source GitHub repo." The GitHub API dates the repo to 2026-06-11, 96 days before the video posted on 2026-09-15.
The product name appears two ways: "Open Executive" in the README heading and OpenExecutive in the repo slug. The organization is Sente Labs (SenteLabsAI, with the site sentelabs.ai), though a demo card in the video reads "Sente Lab."
Which Claude model does Open Executive use by default?
The video says "it defaults using Claude Sonnet 5 under the hood." That matches the README as it stood on the post date. The last README commit before the video, 5bd5f2824c4f on 2026-09-11, sets DEFAULT_MODEL to claude-sonnet-5 for "Executive + most specialists."
That README used three pinned models:
claude-sonnet-5as the default, including the top-level Executive Orchestrator.claude-opus-5for deep reasoning.claude-haiku-4-5for intent routing and background memory extraction.
The default has since moved. The current main README sets DEFAULT_MODEL to claude-sonnet-5-5. Anthropic's release notes date Claude Sonnet 5.5 to September 28, 2026, thirteen days after the video. Anyone who clones the repo now gets the Sonnet 5.5 default.
The model dropdown shown in the video is a motion-graphic mock. The video never shows the product's real interface.
Does Open Executive use Opus for CEO decisions?
Yes, for the decisions that look most like a CEO's job. The same 2026-09-11 README sets DEEP_REASONING_MODEL to claude-opus-5 for "CSO, CFO, GC, Board," with extended thinking. Those are the Chief Strategy Officer, Chief Financial Officer, General Counsel and the board-communications specialist.
The video's punchline points a "CEO tasks" arrow at the Sonnet column. The repo's own routing table does the opposite. Sonnet runs the orchestration layer and the remaining specialists, and Opus handles strategy, financial modeling, legal exposure and board communication. The current README keeps the same split one generation newer: claude-sonnet-5-5 as default and claude-opus-5-5 for "CSO, CFO, GC, Board," with adaptive thinking.
The video also claims Opus has a "Larger context window" than Sonnet. Anthropic's models overview lists a 1M token context window and 128K max output for Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 alike. On the current lineup, context size does not separate Opus from Sonnet.
Does Open Executive really have Harvard MBA-level knowledge?
The video says the repo "reports having Harvard MBA level knowledge," and that is an accurate quote of the README. Nothing in the repository ties the phrase to Harvard. I found no Harvard affiliation, license or curriculum credit in the file tree.
What backs the phrase is a set of hand-written Markdown files. At commit 5bd5f2824c4f, the knowledge/builtin/ folder held 96 Markdown files totaling 666,740 bytes across ten folders: strategy (9), finance (9), product (9), hr (8), legal (8), operations (8), marketing (7), board (6), skills (15) and failures (17). The failures folder holds case studies such as enron.md, theranos.md, ftx-collapse.md and boeing-737-max.md. The README loads them into ChromaDB at startup. That is a specific business reference library, and it is not a Harvard credential.
The repo's one measured quality bar is an eval suite. The README says evals/ holds "29 scenarios covering all 8 domains, scored by claude-opus-4-7 as an LLM-as-judge," rated 1 to 5 on five dimensions, with a CI gate of "≥ 3.5/5 average." That checks consistency against another Claude model. It does not measure executive decision quality.
The README also disagrees with itself on team size. It says "nine specialist AI agents" in one place, "# 9 specialist agents" in another and "any of the ten specialists" in a third, while the GitHub description says "8 specialist agents."
Can AI replace any job, including the CEO?
The video's caption, "AI can replace any job," is the premise of the joke. Three dated sources push back on it.
- METR's trial. A randomized controlled trial by Joel Becker, Nate Rush, Elizabeth Barnes and David Rein (arXiv:2507.09089, July 2025) had 16 experienced developers complete 246 tasks with AI allowed or disallowed. The developers forecast AI would cut completion time by 24% and estimated afterward that it cut time by 20%. The measured result was that "allowing AI actually increases completion time by 19%."
- BLS projections. The Bureau of Labor Statistics Occupational Outlook Handbook, last modified August 27, 2026, counts 3,890,600 top executive jobs in 2025 and projects 5% growth from 2025 to 2035, "faster than the average for all occupations," with about 304,100 openings a year.
- Anthropic's usage data. The Economic Index report "Cadences," published June 26, 2026, found management made up 23% of survey respondents against a 7% employment share but only 4% of sessions. Anthropic reads that as "managers using Claude for tasks other than management itself."
Bottom line
Open Executive is a real Apache-2.0 project that defaulted to Claude Sonnet 5 when the video posted and now defaults to Sonnet 5.5. It routes strategy, finance, legal and board work to Opus, so the "not opus level" joke runs against the repo's own design. The "Harvard MBA-level" line is marketing copy backed by 96 Markdown files.
Sources
- SenteLabsAI/OpenExecutive on GitHub: license, stack, stars and creation date.
- OpenExecutive README at commit 5bd5f2824c4f: the README four days before the post, with the model routing table and eval gate.
- OpenExecutive README on main: the current Sonnet 5.5 and Opus 5.5 routing.
- Models overview, Anthropic docs: current context windows and legacy model list.
- Claude API release notes: launch dates for Sonnet 5.5 and Opus 5.5.
- METR developer productivity study, arXiv:2507.09089: the 19% slowdown result.
- Top Executives, BLS Occupational Outlook Handbook: job count, growth and openings.
- Anthropic Economic Index report: Cadences: management usage figures.
Published September 15, 2026. Based on a TikTok by @alberta.nyc. Full video notes (markdown)