THTutorHero

Everything / Models and research

Megatrend · 16 trends inside · through Sep 2026

Models and research Peaking

On TutorHero's Everything Index, Models and research is peaking on GitHub: 1.17 of every 10,000 new public repos over the last three months (1.1k a month), +64% vs a year ago. The 12-month forecast says edging up (+17%). By the end of 2032, lines in the same state ended between 0.28x and 4.8x of where they started (median 0.82x).

Per 10k new repos1.171.1k new repos a month
Year over year+64%3 months: -4%
Next 12 months+17%80% band -1% to +276%
2032, median of history0.82x90% of past cases: 0.28x to 4.8x
Attention2.62%of new stars, -49% in a year
Jev 2032 call2.2xTop quarter of similar lines
Share on X

Since 2019, and the range to 2032

0.10.31320192020202120222023202420252026202720282029203020312032Range to 2032 (history of similar lines)Stable Diffusion releasedChatGPT launchesGPT-4Llama 2, open weightsModel Context ProtocolDeepSeek-R1Claude Code previewHigh 5.62Jev 2.59Median 0.96Low 0.33

New repos per 10,000 (3-month average, log scale)12-month forecastRange to 2032: middle half and 90% of past similar linesJev's 2032 call

The range is not a forecast of this line: it is what GitHub lines in the same state (grew its share 15% to 100% in a year) did 1 to 5 years later, 2019 to 2026, applied to today's level. Past 5 years it is stretched, not measured. Checked out of time (learned before 2022, scored after), 71% of outcomes landed inside the 90% range, so read it as wide. Open it in the Everything chart.

Week by week

Last week325 new reposOct 12 to Oct 18 · -25% against the field
Last 4 weeks▼ -9%1.1k new repos, against the 4 weeks before and the field
Wikipedia, 7 days11.0k views+3% against the 4 weeks before, all of Wikipedia the same way
AprMayJunJulAugSepOct

Wikipedia views a day, last 8 weeks

Aug 28Sep 4Sep 11Sep 18Sep 25Oct 2Oct 9Oct 16

New public repos with its core tags each week (Monday to Sunday, UTC). The outlined bars, last week included, are still filling in, as tags are often added days after a repo is created, so their height will grow a little. "Against the field" divides the change by what the typical tag did, so a wave that lifts every tag does not count. Hover for the numbers.

Built vs talked about

Hacker News13 stories a month-48% share of HN vs a year ago · titles with "fine-tuning", "transformers", "llm training"
Wikipedia41.7k views a month11x share of Wikipedia vs a year ago · Fine-tuning (deep learning), Transformer (deep learning)
Talk vs buildingTalk and building in stepHacker News -48% vs GitHub new repos +64% over the year, compared with the typical trend

Each line is its own share (of all new GitHub repos, all Hacker News stories, all English Wikipedia views, all new stars on the repos we track), shown as a share of its own peak, so the timing and shape compare even though the units differ. Hover for the counts. Tap a name to hide or show a line.

What it means

Studentslearning it, contributing, getting hiredCommit this year
  • A steady field: Models and research moved +64% in a year, so skills here keep their value without a rush.
  • Start by reading NandhaKishorM/laya (+30.0k stars in 3 months) and rohitg00/ai-engineering-from-scratch (+10.9k stars in 3 months).
  • People to follow: Comfy Org, Unsloth AI, Ilya Grigorik.
  • Jev for someone who wants to contribute or get hired: Commit this year (2.7 of 4).
Founders and builderswhat to build nextTest small
  • Getting crowded: new repos +64% but attention -49%, so each new project competes harder for stars.
  • 20.6 new stars per new repo over 3 months, vs 8.50 for the typical trend: more demand per project than most fields.
  • The 12-month forecast for its share: +17% (Edging up).
  • Talk and building in step: Hacker News -48% vs new repos +64% in a year, close to the typical trend.
  • Jev for a builder: Test small (2.4 of 4).
Investorswhich fields deserve research timeCommit this year
  • History: of past GitHub lines that grew its share 15% to 100% in a year, 49% had a smaller share 3 to 5 years later and 13% at least doubled.
  • Jev's 2032 call: top quarter of similar lines (unsure), which on the base rates means about 2.2x today's share by the end of 2032.
  • Attention: 2.62% of new stars on the repos we track, -49% in a year.
  • Hacker News: 13 stories a month, -48% share in a year; Wikipedia 41.7k views a month, 11x.
  • Jev for an investor: Commit this year (2.6 of 4).
  • Not investment advice: this measures what people build on GitHub, not any company.
Developers choosing toolswhich tools to adoptTest small
  • Most-starred tools here lately: NandhaKishorM/laya (vouched), rohitg00/ai-engineering-from-scratch (vouched).
  • A settled field: favour tools with a long record.
  • Jev for a developer picking tools: Test small (2.3 of 4).

Repos carrying it now

NandhaKishorM/layaNon-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ langua+30.0k stars in 3 months · vouchedrohitg00/ai-engineering-from-scratchLearn AI Engineering! Learn it. Build it. Ship it for others.+10.9k stars in 3 months · vouchedbojieli/ai-agent-book《深入理解 AI Agent：设计原理与工程实践》（李博杰 著）开源主仓库：全书正文、编译版 PDF 与按章配套代码+8.4k stars in 3 months · vouchedHuman-Agent-Society/reefInfrastructure for continually self‑improving agents+7.5k stars in 3 months · newbojieli/ai-infra-book《深入理解 AI Infra：量化分析与系统设计》（李博杰 著）开源书稿：从硬件约束和模型架构出发，量化推导 LLM 推理与训练系统设计。含全书正文、PDF、配套计算工具与实验+5.7k stars in 3 months · vouchedComfy-Org/ComfyUIThe most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the w+5.0k stars in 3 months · vouched

People building it

Comfy OrgComfyUI, a modular diffusion model GUI with graph interface.Influence 94Unsloth AIRunning and training AI models locally with fine-tuning notebooks.Influence 87Ilya GrigorikBuilding merchant+web platform capabilities to supercharge commerce at @Shopify.Influence 86QwenAlibaba Cloud's general-purpose AI modelsInfluence 86Google ResearchOpen-source AI models and research code.Influence 85Sebastian RaschkaBuilding LLMs from scratch and writing machine learning books.Influence 83NVIDIA Research ProjectsNeural graphics, generative image models, and efficient synthesis research.Influence 83Daniel Miessler 🛡️Creating security lists and AI frameworks for human augmentation.Influence 82

Trends inside

Machine learningSteady+72% in a year · 9.9k new repos a monthClaudeSteady11x in a year · 6.8k new repos a monthOpenAISteady+102% in a year · 3.0k new repos a monthPyTorchGrowing+133% in a year · 3.0k new repos a monthDeep learningCooling+20% in a year · 2.6k new repos a monthGeminiGrowing+128% in a year · 2.4k new repos a monthDeepSeekEmerging17x in a year · 2.3k new repos a monthChatGPTAccelerating+257% in a year · 1.3k new repos a monthQwenEmerging11x in a year · 951 new repos a monthTensorFlowFaded-6% in a year · 732 new repos a monthTransformersPeaking+69% in a year · 685 new repos a monthLoRAGrowing+233% in a year · 514 new repos a monthFine-tuningGrowing+79% in a year · 386 new repos a monthKerasFaded+11% in a year · 374 new repos a monthLlamaCooling+167% in a year · 296 new repos a monthLLM trainingCooling-38% in a year · 29 new repos a month

Questions

Is Models and research growing on GitHub?

Yes. Its share of all new public GitHub repos is +64% vs a year ago and -4% over the last three months (Sep 2026). Phase: Peaking.

How many new Models and research repos are created each month?

About 1.1k new public repos a month carry its core tags (three-month average to Sep 2026), or 1.17 of every 10,000 new repos.

What will Models and research look like in 2032?

Nobody knows, and we say how much. Of past GitHub lines that grew its share 15% to 100% in a year, the middle half had between 0.53x and 1.42x of their share 3 to 5 years later, and 49% ended smaller. Jev's call: Top quarter of similar lines.

Is Models and research more talked about than built?

Talk and building in step. Over the year its share of Hacker News stories moved -48% while its share of new GitHub repos moved +64%. New repos grew faster than talk for most AI trends this year, so the label compares it with the typical trend: 1.1 times the talk per builder of the median trend.

Not investment advice. Measurements of what people build and star on GitHub, a statistical forecast, base rates and a model's reading of them. Nothing here is about any company's revenue, valuation or stock.