AI Trajectory
AI Trajectory is a data page on the AI desk of Glassbox Auditing. Six measured vectors for each AI company (departures, release timing, compute, coding, open-marketplace choice, exposure to substitution), each against its peers and its own record, with the rule beside every score.
Last updated: .
What is on this page today?
- Meta Platforms: 3 officer or director departures filed with the SEC in 12 months — 8-K item 5.02, every role, in the 12 months to 7 Nov 2026 · one of the AI Trajectory page's inputs (SEC filings, 7 Nov 2026) META
- Apple: 2 officer or director departures filed with the SEC in 12 months — 8-K item 5.02, every role, in the 12 months to 7 Nov 2026 · one of the AI Trajectory page's inputs (SEC filings, 7 Nov 2026) AAPL
- Alphabet Inc.: 2 officer or director departures filed with the SEC in 12 months — 8-K item 5.02, every role, in the 12 months to 7 Nov 2026 · one of the AI Trajectory page's inputs (SEC filings, 7 Nov 2026) GOOGL
- Microsoft: 2 officer or director departures filed with the SEC in 12 months — 8-K item 5.02, every role, in the 12 months to 7 Nov 2026 · one of the AI Trajectory page's inputs (SEC filings, 7 Nov 2026) MSFT
- Salesforce: 1 officer or director departure filed with the SEC in 12 months — 8-K item 5.02, every role, in the 12 months to 7 Nov 2026 · one of the AI Trajectory page's inputs (SEC filings, 7 Nov 2026) CRM
- Cerebras Systems has gone 44 months without a model that set a new company best on the capabilities index, the longest of 25 companies — months since each company's latest model that beat its own earlier best · one of the AI Trajectory page's inputs; the page's 1-10 scores are a rule-based reading and stay there (Epoch AI Capabilities Index, 7 Nov 2026)
- Cerebras Systems lets other labs use 100% of its AI supercomputer capacity, Microsoft 100% — H100-equivalents of the clusters a company owns that other labs also use · one of the AI Trajectory page's inputs; the page's 1-10 scores are a rule-based reading and stay there (Epoch AI Supercomputers, 7 Nov 2026)
- Google trails SWE-bench's leader Anthropic by 1.0 points; 9 labs are scored — Verified · bash-only mini-SWE-agent (same scaffold for every model) · 1 attempt(s), each company's best model in that set-up · one of the AI Trajectory page's inputs; the page's 1-10 scores are a rule-based reading and… (SWE-bench, 7 Nov 2026)
- IBM had 2.1 researchers leave for every one who joined in 2025, the most of 22 AI companies — OpenAlex author affiliations, by author id · leaving includes moves to academia · one of the AI Trajectory page's inputs; the page's 1-10 scores are a rule-based reading and stay there (OpenAlex, 30 Dec 2025)
Where does the data come from?
Items on this page come from SEC filings; Epoch AI Capabilities Index; Epoch AI Supercomputers; SWE-bench; OpenAlex. Each item links to its own source. All sources, how items are checked.
What does this page lead to?
- Leadership (Companies): The departures vector counts officer and director exits filed under 8-K item 5.02; the same filings name who left and who joined on the Leadership page. (moves the same way; same day; company scale) Source: SEC Form 8-K item 5.02 (director and officer departures and appointments).
- AI Supply Chain (AI): The compute vector reads each company's clusters and capital spending from Epoch AI and SEC filings; that spending is chip, server and data-center revenue for the supply chain. (moves the same way; quarters; sector scale) Source: Epoch AI, AI Supercomputers (CC BY 4.0).
- Competitors (Companies): The substitution vector counts revenue in segments a company's own annual report names AI as a competitive risk for; the same reports' competition sections list the rivals on the Competitors page. (either way, depending on the cause; quarters; company scale) Source: SEC EDGAR filings (10-Q/10-K text).
- Products & Share (Companies): The open-marketplace vector reads each company's share of OpenRouter token volume, a public sample of which models developers pay to run, the usage behind the AI product lines listed companies report. (either way, depending on the cause; months; company scale) Source: OpenRouter model prices.
What feeds this page?
- Benchmarks (AI): Coding-test results are an input to the trajectory's coding vector, so a lab's new coding score changes its gap to the leader there. (moves the same way; days; company scale) Source: SWE-bench leaderboards.
- AI Researchers (AI): Researchers who change employer show up in their OpenAlex affiliations; a lab losing senior authors reads higher on the trajectory's departures vector. (moves the same way; months; company scale) Source: OpenAlex (scholarly works).