Open-weight models have closed to within roughly ten percent of the proprietary frontier, commoditising the model layer and making open-source AI strategically central — yet the honest tension is that the thing being given away is the hardest to monetise, leaving a distinctively low-proximity theme whose value accrues to the cloud and enablement layers, not the model itself.
📈 What changed: Meta doubled down on its proprietary pivot with Muse Spark 1.1 (2026-07-09), a paid agentic-coding model priced per API token — with no accompanying open-weight Llama release, con…
| Indicator | Score | Reading |
|---|---|---|
| Maturity higher is better · high confidence | 71/100 | Mature, widely deployed open weights; architecture converged but 'open' fragmented. |
| Evidence Strength higher is better · high confidence | 61/100 | One verified T2 company-disclosure source (Meta's own Muse Spark announcement) now on file; the rest remains promotional/blog-tier (P3 pip flags this). |
| Commercial Proximity higher is better · medium confidence | 50/100 | The honest tension: real use, murky value capture — the lowest-proximity theme of the cluster. |
| Capital & Policy Support higher is better · medium confidence | 69/100 | Many sovereign-AI strategies favour open weights, but as posture not a spend mandate. |
| Crowding Risk lower is better · medium confidence | 50/100 | No public pure-play; sponsors held for other reasons, so realised crowding is moderate. |
| Reflexivity Risk lower is better · high confidence | 64/100 | Narrative shocks (DeepSeek moments) hit a base of mostly grounded sponsors. |
Every indicator score is computed by the Northstar engine from analyst-set ordinal bands — never hand-written, never stored.
The public page is the summary (backed by 9 cited sources). Full briefing — company map, catalysts, native signals and sources — is inside the app.
Open in the appStart free →The Northstar Signal — what moved across all 350 themes and why, in your inbox. Free, no spam, one-click unsubscribe.