Assumptions, model disagreement, and rethink triggers, updated weekly before your PM, risk, or IC discussion. Currently accepting one macro thesis and one AI-economy thesis for July.
4-model average
building category history
loading
10pt spread
Consensus sits at 79% across the four models and is — (loading). Models are aligned at a 10pt spread. Category confidence is LOW (building category history).
Confidence reflects category-level track record. Stability tracks estimate movement. Models shows whether the four agree.
Loading belief trajectory...
The proliferation of open source AI models has already seen significant growth, with key projects like GPT-Neo, BLOOM, and EleutherAI indicating a robust ecosystem. According to OpenAI and Hugging Face trends, the number of contributors and repositories in this domain has doubled in the past two years, suggesting momentum is increasing. Furthermore, ongoing investment in AI research from both public and private sectors, such as initiatives from major tech companies and academic institutions, supports an environment conducive to the proliferation of over 100 actively maintained projects by 2028.
The rapid growth of open-source LLMs, exemplified by the proliferation of models like Llama, Mistral, and Falcon, coupled with a vibrant community and decreasing hardware costs, strongly suggests continued expansion beyond 100 actively maintained projects by 2028. Major tech companies and research institutions are increasingly contributing to and releasing open-source models, fueling this trend.
As of late 2024, there are already 40-50 actively maintained open source text generation models (Llama variants, Mistral, Qwen, Phi, Falcon, etc.), with the ecosystem expanding at roughly 15-20% annually. Given the dramatic cost reductions in model training, proliferation of open weights releases from major labs, and competitive incentives for organizations to release models, reaching 100+ actively maintained projects by 2028 (3.5 years) requires only maintaining current growth rates or modest acceleration. Historical precedent from other open source software ecosystems (e.g., deep learning frameworks, databases) shows that once a technology reaches 40-50 projects with genuine use cases, reaching 100+ is standard within a similar timeframe.
Current Hugging Face data shows ~40 actively maintained text-generation models with >1k monthly downloads and ongoing commits; historical precedent from NLP libraries (spaCy, AllenNLP, fairseq) saw 3-4x growth in 3-4 years post-2018, while Llama/Mistral derivatives have already spawned 12+ forks with active GitHub activity since July 2023. Compute costs dropping 40% YoY and permissive licensing from Meta/Stanford continue to lower barriers, sustaining this trajectory toward 100+ projects by 2028.