1 February 2026

 OpenAI is retiring several older ChatGPT models (GPT‑4o, GPT‑4.1, GPT‑4.1 mini, o4‑mini, and the GPT‑5 Instant/Thinking variants) from the ChatGPT interface on February 13, 2026, while keeping them available via API for now.llm-stats+1

Daily AI brief – last 24 hours

• OpenAI confirmed the upcoming removal of GPT‑4o, GPT‑4.1, GPT‑4.1 mini, and o4‑mini in ChatGPT on Feb 13, 2026, in addition to earlier-announced GPT‑5 Instant and Thinking retirements.openai+1

• OpenAI states this is to focus UX and engineering on newer GPT‑5.1/5.2 models that incorporate GPT‑4o’s “warm” conversational style and creative ideation strengths, plus more granular tone and style controls.[llm-stats]​

• OpenAI notes that only about 0.1% of daily ChatGPT users still choose GPT‑4o, with “vast majority” already on GPT‑5.2, which they present as a justification for sunsetting 4o.[llm-stats]​

• The same announcement highlights continued work on reducing unnecessary refusals and “overly preachy” answers and on tuning ChatGPT for adults over 18 while adding age‑prediction safeguards for under‑18 users.[llm-stats]​

• Moonshot AI’s Kimi K2.5, a trillion‑parameter, native multimodal model with 256K context and “self‑directed agent swarm” orchestration, is being profiled as one of the most significant recent releases, emphasizing parallel sub‑agents and very long‑context performance.aitoolsguide+1

• Industry trackers show that OpenAI, Anthropic, Google, xAI, and Chinese labs such as Moonshot and Zhipu continue rapid iteration, with new reasoning‑focused and efficiency‑focused LLMs appearing weekly, and Kimi K2.5 highlighted as a major new open‑source option.aitoolsguide+1

• OpenAI’s move fits a wider pattern where providers retire older front‑end models quickly while keeping API endpoints stable longer, nudging everyday users toward current flagship models while minimizing disruption for developers.theregister+1

20+ concrete AI use cases for genealogists

Below are practical, “you can try this today” patterns, framed for an active researcher, teacher, or blogger.

Core research and analysis

1. Brick‑wall brainstorming partner
   Paste a concise research summary (who/where/when, known records, negative searches) and ask an LLM to list additional record types, jurisdictions, or time‑period‑appropriate sources you may have missed (e.g., poor law, tax, militia, chancery).denyseallen.substack+1

2. Research plan drafting
   Feed the AI a research question and current evidence and ask for a prioritized, source‑citation‑oriented plan (with repositories, years, and record groups) that you then refine against the Genealogical Proof Standard.[denyseallen.substack]​

3. Hypothesis generation from deeds and legal prose
   Paste your own transcriptions of complex deeds or settlements and ask the model to (a) list all named parties with inferred roles, (b) outline possible relationship hypotheses, and (c) flag conflicts you should test, explicitly labeling all output as unproven.[blog.dnapainter]​

4. Timeline and correlation grids
   Give the AI a set of extracted facts (dates, places, events, source notes) and ask it to produce a chronological table with inferred gaps or conflicts you need to address, then export or copy this into your own research log.[denyseallen.substack]​

5. Negative‑evidence articulation
   Describe exhaustive searches you’ve already done (collections, date ranges, jurisdictions) and have the AI help you phrase clear negative‑evidence statements for your proof arguments or reports.[denyseallen.substack]​

Working with documents

6. Handwriting transcription (drafts)
   Use AI handwriting/OCR tools (e.g., platform transcription, Gemini‑powered tools, or site‑provided AI transcription) on 18th–20th‑century deeds, mortgages, probate records, parish registers, or court minutes, then compare carefully against the images and correct errors for your final citation.njstatelib+1

7. Summarizing long documents
   Once transcribed, paste long wills, chancery cases, or narrative obituaries into an LLM and ask for: “A structured summary listing heirs, property, places, and conditions,” while keeping your own copy as the authoritative version.dnapainter+1

8. Index‑like extractions
   For a run of similar documents you’ve transcribed (e.g., a series of deeds), ask the AI to build a table of names, roles, dates, locations, and relationships you can then import into a spreadsheet or database.[blog.dnapainter]​

9. Language translation of records
   Use AI translation to get a working English rendition of records in Latin, German, Spanish, Scandinavian languages, or others, and then consult specialist references or human experts to confirm key genealogical terms.[njstatelib]​

10. Newspaper OCR cleanup
    Paste messy OCR from old newspapers into an LLM and ask it to normalize spelling, re‑insert paragraph breaks, and identify people, places, and events related to your research question.njstatelib+1

DNA and inferential work

11. Explaining DNA results in plain language
    Summarize the key facts from a DNA match cluster (centimorgans, tree hints, shared locations) and ask the model to explain likely relationship ranges and conventional next steps, noting you will verify everything against testing‑company tools and human expertise.[blog.dnapainter]​

12. Organizing hypothesized genetic networks
    Provide anonymized, high‑level descriptions of match groups (shared segments, ancestral locations, surname clusters) and ask the AI to outline alternative relationship scenarios you might diagram in Lucidchart or DNA Painter.[blog.dnapainter]​

Writing and publishing

13. Drafting ancestor biographies
    Give the AI your fully cited notes for one ancestor (facts only, with your own wording) and ask for a narrative draft in a specified voice (e.g., neutral academic, magazine‑style) that you then revise, fact‑check, and annotate with your own citations.[njstatelib]​

14. Turning research logs into blog posts
    Paste a cleaned‑up research log and ask for 2–3 possible blog‑post outlines, each with headings, sidebars (maps, timelines), and call‑outs for teaching points about methodology or record sets.aigenealogyinsights+1

15. Title, subtitle, and SEO idea generation
    Give the AI your draft post and ask for alternate titles, meta descriptions, and short social‑media blurbs tailored for genealogists and family historians, then select and adapt what fits your voice.aitoolsguide+1

16. Checking narrative flow and audience level
    Ask the model to analyze a draft chapter or lesson and suggest edits to make it clearer for beginners, or more rigorous for advanced researchers, without altering factual content.[denyseallen.substack]​

17. Creating teaching handouts and checklists
    From your syllabus or talk outline, have AI generate a one‑page checklist, glossary, or worksheet (for example, “Pre‑research checklist for rural Southern U.S. ancestors, 1850–1900”) that you then brand, proof, and publish as a PDF.aigenealogyinsights+1

Visuals, photos, and presentation

18. Photo enhancement and colorization
    Use AI photo tools on copies of images to colorize, sharpen, or repair damaged family photos, always preserving the unedited originals and labeling AI‑modified versions clearly.[njstatelib]​

19. Story‑supporting image prompts
    For blog posts or teaching slides, ask a model to suggest period‑appropriate, non‑specific image ideas (e.g., “generic 1880s Midwestern farm kitchen”) you can then generate in an image tool, avoiding depictions of real ancestors to prevent confusion.[aitoolsguide]​

Teaching, and mentoring uses

20. Lesson outlines for genealogy societies, webinars or community groups
    Provide your historical framework and ask AI for a 30‑ or 60‑minute lesson outline that uses family‑history examples to illustrates methods, memory, or topics which you then adjust for your  context.aitoolsguide+1

21. Q&A preparation for workshops
    Compile typical attendee questions (“How do I start with no living relatives?” “What about closed adoptions?”) and have AI help you draft concise, sensitive, and methodologically sound responses you can adapt into a FAQ sheet.aigenealogyinsights+1

22. Case‑study packaging
    Take one of your solved cases and ask the model to help structure it into a teachable case study: background, research question, sources consulted, conflicts encountered, resolution, and lessons learned.aigenealogyinsights+2

23. Email/newsletter drafting for a genealogy society
    Feed the AI bullet‑point updates (events, new resources, calls for volunteers) and request a short, conversational newsletter draft, then revise to keep your personal voice and ethical boundaries intact.aitoolsguide+1

24. Prompt libraries for recurring tasks
    As you experiment, save and refine prompts that work well for you (e.g., “evidence summary prompt,” “deed‑analysis prompt,” “ancestor‑bio prompt”) and ask the model to help standardize them into a one‑page reference sheet you can share with students.aigenealogyinsights+1

For immediate experimentation this week, you might choose one stubborn brick wall, one dense deed or will, and one half‑finished blog post, and run each through a carefully instructed AI workflow like those above, documenting what genuinely helps—and what you would never outsource.