Skip to report content 
+--------------------------------------------------------------+
| OSS AI STACK MAP :: SNAPSHOT REPORT                          |
| SOURCE: data/run-2026-07-30-repaired-v3                      |
| LENS: major, active, public OSS AI repos on GitHub           |
+--------------------------------------------------------------+

What major open source AI repos actually use.

This report reads directly from data/run-2026-07-30-repaired-v3 and summarizes the current stack choices across the project’s final GitHub AI set.

Study frame: GitHub-only, public, non-fork, non-archived, active within 1 month, and at least 1,000 stars. Published stack edges come from manifests, SBOMs, bounded import fallback, repo identity, and explicitly approved README fallback when an included repo would otherwise remain unmapped.

Read the explainer
+ Discovered repos
1,504
Candidate universe collected from GitHub discovery queries and manual seeds.
+ Final major AI repos
1,043
69.3% of discovered repos survive the full inclusion filter.
+ Normalized technology edges
4,713
994 repos map to at least one of 77 tracked technologies.
+ Median final repo stars
4,980
P75 is 15,275 stars across the published final set.
+ main findings

The modern OSS AI stack is provider-first, orchestration-heavy, and surprisingly multi-vendor.

[finding]

Providers sit at the center

Provider SDKs account for 25.0% of all normalized edges, and OpenAI SDK appears in 504 final repos. The provider layer is the most common and the most connective part of the stack.

[finding]

Multi-provider stacks are common

299 of 994 technology-mapped repos (30.1%) use at least two tracked providers. The most common provider pairing is Anthropic SDK plus OpenAI SDK in 233 repos. That is followed by Google GenAI SDK plus OpenAI SDK in 194 repos. Major OSS projects are not clustering around a single vendor.

[finding]

The ecosystem is backend-heavy

Training, orchestration, providers, and retrieval dominate the graph. Evaluation and observability remain comparatively thin, with only 28 guardrail/eval edges and 67 observability edges.

+ stack profile

A compact read on the default stack shape

Inference from the aggregate counts: the modal major OSS AI repo in this snapshot is organization-owned, Python-first, anchored on a provider SDK, often layers in Hugging Face training tools, and then adds orchestration, retrieval, and a lightweight UI shell.

[who ships it]
Organization
785 / 75.3%
User
258 / 24.7%
[language mix]
Python
501 / 48.0%
TypeScript
284 / 27.2%
Go
64 / 6.1%
Rust
53 / 5.1%
JavaScript
29 / 2.8%
+ entity mapping

Who appears to steward repos, and which companies sit behind tracked technologies

This layer is separate from stack normalization. Repo steward mapping currently uses curated exact repo-name and GitHub org matches only. Technology vendor mapping is curated in config and should be read as product stewardship, not proof that every adopting repo is company-backed.

+ Tracked entities
24
Curated company, startup, foundation, and individual entities loaded from the entity registry.
+ Steward-mapped repos
64
6.1% of final repos map to a steward entity via repo-name or GitHub org evidence.
+ Vendor-mapped repos
836
80.2% of final repos use at least one technology tied to a curated vendor entity across 34 tracked technologies.
[top repo stewards]
Hugging Face
company
12 / 1.1%
LangChain
startup
11 / 1.1%
Google
company
10 / 1.0%
E2B
startup
5 / 0.5%
OpenAI
company
3 / 0.3%
Qdrant
startup
3 / 0.3%
[top technology vendors]
OpenAI
2 mapped technologies
523 / 50.1%
Hugging Face
5 mapped technologies
403 / 38.6%
Anthropic
1 mapped technologies
303 / 29.0%
LangChain
5 mapped technologies
211 / 20.2%
Google
1 mapped technologies
198 / 19.0%
Vercel
2 mapped technologies
125 / 12.0%
[entity types]
startup
16
company
8
[repo steward confidence]
high
64
+ coverage and method

Evidence tiers, validation audit, and where the map still undercounts

+ Direct-supported repos
994
Final repos with manifest, SBOM, import, or repo-identity evidence and no dependence on README-only fallback for coverage.
+ Approved fallback repos
0
Additional final repos mapped through explicitly approved README-only evidence.
+ README-only repos
0
Low-confidence fallback repos that remain explicit and separately inspectable in the publication artifact.
+ Pending README candidates
121
Automatically detected README matches awaiting an explicit curator decision.
[1] Evidence tiers are now explicit

1,032 final repos have manifests and 816 have SBOM dependency evidence. 12 repos map only via canonical repo identity, 0 combine direct evidence with fallback signals, and 0 remain README-only.

[2] Normalization still leaves gaps

49 included repos (4.7%) have no normalized technology edge, so graph-like analysis describes the mapped subset, not the entire final population.

[3] Validation state is explicit

0 repos were judge-reviewed and 0 judge overrides were applied in this snapshot. The published set remains rule-driven. A judge-backed validation audit was not run for this snapshot; no false-positive estimate is available.

[4] Collection provenance is measurable

0 repos record a material fetch or parse failure. 1,504 legacy collection outcomes remain unknown because this repaired snapshot predates per-source outcome tracking.

[evidence profiles]
direct only
982
unmapped
49
repo identity only
12
[validation sample by segment]
No validation audit segment data in this snapshot.
+ research gaps

Where normalization still misses stack evidence

+ Missing-edge repos
49
Final repos that still have no normalized technology edge.
+ Unmapped dep gaps
45
Missing-edge finals that do have dependency evidence, but none of it normalizes yet.
+ No-dependency gaps
4
Missing-edge finals that have no dependency evidence at all.
+ AI-specific prefixes
10
High-signal unresolved package families that still look AI-stack specific.
+ Commodity prefixes
10
Generic tooling and ecosystem noise separated from the research backlog.
+ Vendor-like repos
10
Repos that look vendor-related but are not mapped to a canonical identity.
[ai-specific unmatched prefixes]
langchain
736
nvidia
669
llama
484
datasets
429
@langchain/
291
agent
232
[largest missing-edge repos]
moltis-org/moltis
unmapped dependency evidence • unmatched deps 1,980
2,797 stars
agentgateway/agentgateway
unmapped dependency evidence • unmatched deps 1,750
4,122 stars
thClaws/thClaws
unmapped dependency evidence • unmatched deps 844
1,183 stars
mock-server/mockserver-monorepo
unmapped dependency evidence • unmatched deps 709
4,929 stars
lioensky/VCPToolBox
unmapped dependency evidence • unmatched deps 652
2,218 stars
nolabs-ai/nono
unmapped dependency evidence • unmatched deps 619
3,322 stars
[commodity and tooling backlog]
typescript
4,073
serde
3,853
github
3,367
@img/
2,868
windows
2,598
tokio
2,442
+ benchmark recall

How well the map covers the current benchmark panel, holdout set, and negative controls

+ Positive entities
49
Tracked OSS AI stack entities used for recall measurement across this snapshot.
+ Negative controls
10
Discovered candidate repos that should stay out of the final set and act as a precision guardrail.
+ Holdout entities
12
Positive entities reserved as a holdout slice instead of the main tuning panel.
+ Failed thresholds
0
Recall metrics currently below configured minimums.
+ Prioritized gaps
7
Benchmarks needing the next registry or discovery fixes.
+ Total benchmark entries
59
Combined positives and negatives now tracked by the benchmark report.
[coverage rates]
Repo discovered
100.0%
Repo included
98.0%
Identity mapped
95.9%
Third-party adoption
87.8%
Dependency evidence
98.0%
Negative excluded
100.0%
Holdout discovered
100.0%
[top benchmark gaps]
LanceDB
repo included but canonical repo identity is not mapped
score 80
Daytona
benchmark repo discovered but not included
score 70
Haystack
dependency evidence exists but no third-party adoption edge was created
score 60
DingoDB
no dependency evidence found for benchmark entity
score 40
AutoGen
benchmark repo discovered only via exact seed
score 30
+ robustness checks

How much the topline changes under stricter evidence and rule-only views

+ Rule-only final repos
1,043
Final-set size if the project uses raw rule outputs with no judge adjustment.
+ Judge-adjusted final repos
1,043
Published final-set size after hardening and validation overrides are applied.
+ Judge-changed finals
0
Repos whose final-set status differs between rule-only and published judge-adjusted views.
+ Direct-supported repos
994
Mapped repos using direct evidence or repo identity only.
+ Approved fallback lift
0
Additional mapped repos recovered through explicitly approved README-only evidence.
+ Temporal delta
0
Change in final repos relative to the baseline snapshot used for this validation pass.

Rule-only yields 1,043 final repos. Judge adjustment yields 1,043. Direct-only evidence maps 994 repos, and approved fallback lifts that to 994. No temporal baseline comparison is available for this snapshot.

+ technology discovery

Post-filtered canonical candidates from the raw unmatched package graph

+ Raw candidate families
85
Unmatched technology families inferred from scraped dependency evidence before curation filters.
+ Graph nodes
85
Package-family nodes in the projected co-usage graph.
+ Filtered candidates
0
Canonical vendor, product, or package-family candidates that survive the registry filter.
No post-filtered canonical discovery candidates available for this snapshot.
+ registry suggestions

Canonical vendor, product, and package-family suggestions

+ Suggestions
0
Candidates produced after filtering already-covered families and suppressing abstract capability labels.
+ LLM reviewed
0
Optional OpenAI registry reviews attached to candidate suggestions.
+ Shown here
0
Top suggestion rows displayed in this report.
No registry suggestions available for this snapshot.
+ modern ai stack layers

Where normalized stack usage concentrates

These cards rank categories by normalized repo-tech edges, not by architectural importance. Each card now shows both edge share and repo prevalence across the 1,043 final repos. Very thin categories are summarized separately below.

+ Model frameworks and HF stack
1,193
normalized repo-tech edges
[25.3% of edges]
[431 repos / 41.3%]
Transformers
314 / 30.1%
PyTorch
270 / 25.9%
Hugging Face Hub
195 / 18.7%
Tokenizers
141 / 13.5%
Accelerate
132 / 12.7%
+ Providers and access
1,177
normalized repo-tech edges
[25.0% of edges]
[620 repos / 59.4%]
OpenAI SDK
523 / 50.1%
Anthropic SDK
303 / 29.1%
Google GenAI SDK
198 / 19.0%
LiteLLM
153 / 14.7%
+ Orchestration and agents
789
normalized repo-tech edges
[16.7% of edges]
[268 repos / 25.7%]
LangChain ecosystem
LangChain 199, LangChain OpenAI Integration 138, LangChain Anthropic Integration 64, LangChain Google GenAI Integration 47, LangGraph 105
211 / 20.2%
LlamaIndex
42 / 4.0%
OpenAI Agents
41 / 3.9%
CrewAI
34 / 3.3%
Google ADK
26 / 2.5%
+ Protocols and developer SDKs
608
normalized repo-tech edges
[12.9% of edges]
[531 repos / 50.9%]
Model Context Protocol
482 / 46.2%
Vercel AI SDK
124 / 11.9%
code2prompt
1 / 0.1%
Grafbase
1 / 0.1%
+ Retrieval and vector storage
315
normalized repo-tech edges
[6.7% of edges]
[176 repos / 16.9%]
Chroma
74 / 7.1%
Qdrant
71 / 6.8%
pgvector
46 / 4.4%
LanceDB
45 / 4.3%
Weaviate
39 / 3.7%
+ Serving and local runtimes
228
normalized repo-tech edges
[4.8% of edges]
[177 repos / 17.0%]
Ollama
93 / 8.9%
vLLM
52 / 5.0%
Ray Serve
46 / 4.4%
llama.cpp
22 / 2.1%
SGLang
12 / 1.2%
+ UI and app frameworks
150
normalized repo-tech edges
[3.2% of edges]
[130 repos / 12.5%]
Gradio
74 / 7.1%
Streamlit
66 / 6.3%
Chainlit
10 / 1.0%
+ Sandbox and isolated execution
111
normalized repo-tech edges
[2.4% of edges]
[81 repos / 7.8%]
E2B
45 / 4.3%
Daytona
29 / 2.8%
Modal
22 / 2.1%
Vercel Sandbox
7 / 0.7%
Cloudflare Containers
5 / 0.5%
+ Observability
67
normalized repo-tech edges
[1.4% of edges]
[63 repos / 6.0%]
Langfuse
43 / 4.1%
Logfire
16 / 1.5%
Arize Phoenix
4 / 0.4%
Weave
2 / 0.2%
Evidently
1 / 0.1%
+ Browser and computer use infra
42
normalized repo-tech edges
[0.9% of edges]
[36 repos / 3.5%]
Browserbase
24 / 2.3%
Browser Use
11 / 1.1%
Hyperbrowser
3 / 0.3%
Steel Browser
2 / 0.2%
Browserless
1 / 0.1%
+ Evaluation and guardrails
28
normalized repo-tech edges
[0.6% of edges]
[26 repos / 2.5%]
Ragas
12 / 1.2%
DeepEval
7 / 0.7%
Guardrails
4 / 0.4%
Promptfoo
3 / 0.3%
NeMo Guardrails
2 / 0.2%
+ thinly tracked layers

Visible, but not broad enough for a primary card

Runtime and agent deployment
5 normalized repo-tech edges across 5 repos. Thin categories stay listed here until coverage is broad enough for a full card.
Cloudflare Agents 5
[0.1% edges • 0.5% repos]
+ top technologies

The most repeated building blocks

OpenAI SDK
523 / 50.1%
Model Context Protocol
482 / 46.2%
Transformers
314 / 30.1%
Anthropic SDK
303 / 29.1%
PyTorch
270 / 25.9%
LangChain ecosystem
LangChain 199, LangChain OpenAI Integration 138, LangChain Anthropic Integration 64, LangChain Google GenAI Integration 47, LangGraph 105
211 / 20.2%
Google GenAI SDK
198 / 19.0%
Hugging Face Hub
195 / 18.7%
LiteLLM
153 / 14.7%
Tokenizers
141 / 13.5%
+ provider and segment mix

Who dominates, and what gets built

[provider prevalence]
OpenAI SDK
504 / 48.3%
Anthropic SDK
278 / 26.7%
Google GenAI SDK
233 / 22.3%
[primary segments]
Serving runtime
351 / 33.7%
Training and finetuning
210 / 20.1%
Agent application
152 / 14.6%
Orchestration framework
132 / 12.7%
Vector and retrieval infra
83 / 8.0%
AI developer tool
47 / 4.5%
+ repeated combinations

The strongest co-usage patterns

Model Context Protocol :: OpenAI SDK
Shared by major OSS AI repos in the normalized graph.
[271]
Anthropic SDK :: OpenAI SDK
Shared by major OSS AI repos in the normalized graph.
[254]
PyTorch :: Transformers
Shared by major OSS AI repos in the normalized graph.
[230]
Anthropic SDK :: Model Context Protocol
Shared by major OSS AI repos in the normalized graph.
[204]
OpenAI SDK :: Transformers
Shared by major OSS AI repos in the normalized graph.
[190]
Google GenAI SDK :: OpenAI SDK
Shared by major OSS AI repos in the normalized graph.
[170]
+ at a glance

A few stable signals from the snapshot

+ Serious repos
1,319
87.7%
+ AI-relevant repos
1,045
69.5%
+ Repos with normalized techs
994
95.3%
+ Median mapped tech count
3
Across the technology-connected subset only.
+ graph structure

Which technologies sit at the center of the mapped stack

These visuals summarize the technology-connected subset of the final population. Eigenvector highlights the core hubs, betweenness isolates bridge technologies, repo degree shows stack breadth per mapped repo, and category mixing shows which layers of the stack actually co-occur.

[top eigenvector technologies]
High-eigenvector technologies are not just common. They sit next to other highly connected technologies and define the graph’s center of gravity.
Technology eigenvector centralityRanks technologies by eigenvector centrality in the projected co-usage graph.00.10.30.4OpenAI SDK0.4Model Context Protocol0.3Anthropic SDK0.3Transformers0.3PyTorch0.3LangChain0.2Hugging Face Hub0.2Google GenAI SDK0.2LiteLLM0.2LangChain OpenAI Integration0.2Tokenizers0.2Accelerate0.2
View eigenvector data
TechnologyEigenvector centrality
OpenAI SDK0.4231
Model Context Protocol0.3326
Anthropic SDK0.3051
Transformers0.3049
PyTorch0.2691
LangChain0.2377
Hugging Face Hub0.2282
Google GenAI SDK0.2253
LiteLLM0.1904
LangChain OpenAI Integration0.1811
Tokenizers0.1719
Accelerate0.1701
[betweenness vs prevalence]
Points higher on the chart bridge otherwise different tool combinations. Point size reflects weighted co-occurrence strength.
Technology prevalence and betweennessCompares repository prevalence with betweenness centrality; point size represents weighted graph strength.0261.55230.0000.3590.717Hugging Face HubLangChainModel Context ProtocolQdrantTransformersOpenAI SDKVercel AI SDKPyTorchRepo countBetweenness centrality
View scatter plot data
TechnologyRepo countBetweennessWeighted strength
OpenAI SDK5230.71732,999
Transformers3140.18592,034
Model Context Protocol4820.07322,190
LangChain1990.05001,649
Vercel AI SDK1240.0253709
Hugging Face Hub1950.02531,470
Qdrant710.0242742
PyTorch2700.00381,774
Anthropic SDK3030.00231,932
LangChain OpenAI Integration1380.00111,251
PEFT890.0005753
Milvus360.0004346
Browserbase240.0004211
LangChain Google GenAI Integration470.0004410
LiteLLM1530.00031,292
Daytona290.0002280
Streamlit660.0002591
Modal220.0002234
Accelerate1320.00021,092
LangGraph1050.00011,047
Tokenizers1410.00011,122
vLLM520.0000458
Chroma740.0000786
Langfuse430.0000347
E2B450.0000343
Weaviate390.0000380
pgvector460.0000358
code2prompt10.00000
Gradio740.0000620
LlamaIndex420.0000430
Browserless10.00000
Arize Phoenix40.000048
Ray Serve460.0000355
Google ADK260.0000286
Promptfoo30.000021
OpenAI Agents410.0000452
DeepSpeed320.0000237
Semantic Kernel80.0000102
Helicone10.000012
Guardrails40.000038
Chainlit100.0000112
Browser Use110.000089
DSPy110.0000130
DeepEval70.000073
LanceDB450.0000424
CrewAI340.0000451
Weave20.000023
Grafbase10.00000
SGLang120.0000125
Logfire160.0000134
TRL190.0000176
Vercel Sandbox70.000045
LangChain Anthropic Integration640.0000612
Hyperbrowser30.000025
DingoDB10.00000
PrimeIntellect Verifiers10.000014
llama.cpp220.0000240
TGI10.00007
AutoGen60.000095
Haystack10.00001
smolagents90.0000115
Cloudflare Agents50.000025
BentoML20.000010
Ollama930.0000700
NeMo Guardrails20.000026
Google GenAI SDK1980.00001,408
Cloudflare Containers50.000024
Evidently10.00000
Mastra180.0000229
Runloop30.000024
Instructor250.0000302
Ragas120.0000132
PydanticAI150.0000212
Notte10.00006
Steel Browser20.000014
Databend30.000010
[repo degree distribution]
This is the distribution of tracked technologies per mapped repo, not the full final set including no-edge repos.
Repository technology-count distributionShows how many mapped repositories use each range of tracked technology counts.123456-78-1011-1516+0124248
View distribution data
Tracked technologies per repoRepo count
1248
2148
3112
495
581
6-7104
8-10105
11-1573
16+28
[category mixing heatmap]
Weighted co-occurrence between the biggest stack categories. Darker cells indicate heavier cross-category coupling in the technology projection.
Technology category co-usageShows weighted co-occurrence counts between the largest technology categories.TrainingProvidersOrchestrationai_developRetrievalServingUIsandbox_anTraining1,7231,6191,119486572711422141Providers1,6197931,640930676387256232Orchestration1,1191,6401,342672615265272182ai_develop4869306727721712974112Retrieval57267661521725513911288Serving711387265129139608029UI42225627274112802025sandbox_an14123218211288292545
View category mixing data
Category ACategory BWeighted co-occurrence
Model frameworks and HF stackModel frameworks and HF stack1,723
Model frameworks and HF stackProviders and access1,619
Model frameworks and HF stackOrchestration and agents1,119
Model frameworks and HF stackProtocols and developer SDKs486
Model frameworks and HF stackRetrieval and vector storage572
Model frameworks and HF stackServing and local runtimes711
Model frameworks and HF stackUI and app frameworks422
Model frameworks and HF stackSandbox and isolated execution141
Providers and accessProviders and access793
Providers and accessOrchestration and agents1,640
Providers and accessProtocols and developer SDKs930
Providers and accessRetrieval and vector storage676
Providers and accessServing and local runtimes387
Providers and accessUI and app frameworks256
Providers and accessSandbox and isolated execution232
Orchestration and agentsOrchestration and agents1,342
Orchestration and agentsProtocols and developer SDKs672
Orchestration and agentsRetrieval and vector storage615
Orchestration and agentsServing and local runtimes265
Orchestration and agentsUI and app frameworks272
Orchestration and agentsSandbox and isolated execution182
Protocols and developer SDKsProtocols and developer SDKs77
Protocols and developer SDKsRetrieval and vector storage217
Protocols and developer SDKsServing and local runtimes129
Protocols and developer SDKsUI and app frameworks74
Protocols and developer SDKsSandbox and isolated execution112
Retrieval and vector storageRetrieval and vector storage255
Retrieval and vector storageServing and local runtimes139
Retrieval and vector storageUI and app frameworks112
Retrieval and vector storageSandbox and isolated execution88
Serving and local runtimesServing and local runtimes60
Serving and local runtimesUI and app frameworks80
Serving and local runtimesSandbox and isolated execution29
UI and app frameworksUI and app frameworks20
UI and app frameworksSandbox and isolated execution25
Sandbox and isolated executionSandbox and isolated execution45
+ community structure

How the technology graph breaks into stack families

Greedy modularity on the technology projection found 8 communities with modularity 0.1286. Lower modularity means the graph is still heavily cross-linked, so these are useful stack families rather than cleanly isolated islands.
[community 1]

Provider And Orchestration Layer

The dominant application-layer cluster where provider SDKs, orchestration frameworks, guardrails, and observability tools co-occur.
share 51.3%
39 technologies
top categories
Orchestration (14)sandbox_and_isolated_execution (6)Eval (4)
top technologies
Anthropic SDKArize PhoenixAutoGenBrowser UseBrowserbaseCloudflare Agents
exemplar repos
ComposioHQ/composiocomet-ml/opikArize-ai/phoenixawslabs/agentcore-samplesag-ui-protocol/ag-ui
[community 2]

Training And Inference Core

Model-training and inference-runtime technologies clustered around finetuning, serving, and heavyweight model execution.
share 22.4%
17 technologies
top categories
Training (9)Serving (6)UI (1)
top technologies
AccelerateBentoMLDeepSpeedGradioHeliconeHugging Face Hub
exemplar repos
mudler/LocalAIelizaOS/elizamicrosoft/LMOpsray-project/rayOpenBMB/MiniCPM
[community 3]

Retrieval And App Surface

Vector storage, retrieval plumbing, and lightweight app frameworks that often sit at the presentation edge of AI systems.
share 19.7%
15 technologies
top categories
Retrieval (7)Orchestration (3)UI (2)
top technologies
ChainlitChromaDatabendInstructorLanceDBMilvus
exemplar repos
microsoft/semantic-kernellangroid/langroiddynamiq-ai/dynamiqlanggenius/difymicrosoft/autogen
+ run-over-run comparison

What changed since the 2026-03-25 snapshot

This section compares the current 2026-07-30 snapshot against the earlier published 2026-03-25 snapshot (run-2026-03-31-publication-v8). Shares are of each run's own final repo set; the scorecard marks benchmark recall and coverage metrics as improved or regressed. The LLM-judge hardening layer ran in one run but not the other; the 2026-07-30 pass is rule-only, so its serious/AI-relevance filtering relies on rule scores alone — the “Judge-reviewed repos” scorecard row reflects this asymmetry, not a real coverage change.

+ Δ Discovered repos
-21
1,525 → 1,504 since 2026-03-25.
+ Δ Final repos
+60
983 → 1,043 in the published set.
+ New entrants
433
Repos that joined the final set vs. 2026-03-25.
+ Dropped repos
373
Repos that left the final set vs. 2026-03-25.
[quality scorecard]
3 improved · 5 regressed · 4 unchanged
metric
2026-03-25
2026-07-30
Δ
Benchmark repo discovery
100.0%
100.0%
+0.0 pp
Benchmark repo discovery by broad search
91.8%
93.9%
+2.0 pp
Benchmark repo inclusion
100.0%
98.0%
-2.0 pp
Benchmark repo identity mapping
98.0%
95.9%
-2.0 pp
Benchmark third-party adoption
87.8%
87.8%
+0.0 pp
Benchmark dependency evidence
98.0%
98.0%
+0.0 pp
Benchmark failed thresholds
0
0
+0
Final repos
983
1,043
+60
Judge-reviewed repos
651
0
-651
Final repos missing tracked edges
0
49
+49
Missing-edge repos with unmapped dependency evidence
0
45
+45
Missing-edge repos with no dependency evidence
0
4
+4
[biggest technology share movers]
Tokenizers
-11.6 pp
Hugging Face Hub
-9.7 pp
Accelerate
-8.6 pp
Transformers
-8.3 pp
OpenAI SDK
-7.4 pp
PyTorch
-7.2 pp
[top new final-set entrants]
affaan-m/ECC
235,943 ★
Significant-Gravitas/AutoGPT
185,745 ★
Graphify-Labs/graphify
98,781 ★
JuliusBrussee/caveman
94,502 ★
DietrichGebert/ponytail
92,046 ★
nexu-io/open-design
82,671 ★
netdata/netdata
79,922 ★
koala73/worldmonitor
76,734 ★
Full detail, including dropped repos, provider and segment shifts, and edge churn, is in docs/run-comparison-2026-03-25-vs-2026-07-30.md.