B4 Index / The Continuum / July 22, 2026
The Continuum
Week of July 15 to July 22, 2026 · 3 score movements, 2 capability signals, 11 funding events, 1 papers
1,850
Categories tracked
3
Score movements
$4.2B
Capital tracked
2
Capabilities confirmed
1
Papers reviewed
The prior assessment credited mature open-source options and documented production self-builds. Research found neither holds up in 2026: Grouparoo's cloud was discontinued after the Airbyte acquisition, Multiwoven has little independently demonstrated production adoption, dbt positions itself as the modeling layer rather than the sync runtime, and no named company publicly documents running a home-grown reverse-ETL system in production.
JUL 20
Score movements
The prior 4 read the maturity of the open components as evidence that teams were running them.
Corrected an evidence error in the prior score rather than a market move: the prior 4 counted Fonteva, Nimble AMS, MemberVerse, and HubSpot Member Center as independent self-builds, but these are commercial AMS products.
Funding
11| Company | Round | Categories | Announced | Amount |
|---|---|---|---|---|
| AtomsSource | Venture | JUL 22 | $1.7B | |
| AtomsSource | Equity round (unlabeled) | JUL 22 | $1.7B | |
| MeshySource | Series B | JUL 21 | $400M | |
| GlowSource | Series A | JUL 22 | $180M | |
| Candid HealthSource | Series D | Revenue Cycle Management (RCM) | JUL 22 | $120M |
| ArrakisSource | Series A | AI Agent Frameworks & Orchestration | JUL 22 | $38M |
| NaturalSource | Series A | Payments Processing+1 | JUL 20 | $30M |
| GrittSource | Series A | JUL 21 | $26M | |
| InfinitySource | Seed | Multi-LoRA Adapter Serving & Inference Optimization+1 | JUL 20 | $15M |
| CascadeSource | Seed | Construction Project Intelligence & Bid Lead Data | JUL 22 | $3.5M |
| CascadeSource | Seed | Construction CRM & Pursuit Management+1 | JUL 21 | $3.5M |
Capability signals
2Papers
1Loading an agent skill does not keep its requirements active in long contexts — a code-audit task passed 8/10 runs in an ~11k-character clean context but only 3/10 at 299k characters (trend-level, p=0.0698), while a detailed external checklist passed 10/10 versus 5/10 for generic self-checking (p=0.0325); a second task passed all conditions, so no universal context-length threshold is supported.
When and How Context Rot Appears in Coding Agents: A White-Box Study of Agent Skills in Code AuditingWhat it means For long agent sessions, externalize compliance into explicit checklists the agent must tick against rather than trusting a loaded skill or self-review to stay active; small run counts, so treat effect sizes as indicative.
arXiv preprint · JUL 20
Every row here cleared the pipeline's verification before it published, and the research is free. The full database and the score updates behind it are in a B4 subscription.