Muskogee County Genealogical Society

06/03/2026

ChatGPT found several remote or remote-possible genealogy-related postings/contract paths that are interesting if you are looking for a job in the field of genealogy research. date: June 3, 2026.

06/01/2026

Based on search results, here's the latest information about migrating legacy AI models, with specific context about recent retirements (o3, GPT-4.5, Gemini 2.0 models) and practical migration workflows.

What follows is a a detailed step-by-step guide for researchers, particularly genealogists, created by Perplexity AI. For more information about the latest updates see https://last24zotero.blogspot.com/2026/06/1-june-2026.html

# # Step-by-Step Guide: Migrating Legacy AI Models for Researchers

With OpenAI's o3 retiring August 26, 2026, GPT-4.5 retiring June 27, 2026, and Google's Gemini 2.0 Flash models discontinued today (June 1, 2026), researchers must migrate workflows to avoid service disruptions. This guide provides a structured migration framework applicable to genealogical research, historical analysis, and archival workflows.

# # Phase 1: Audit Your Current AI Dependencies

# # # Identify affected workflows

Document every place you use retiring models:

- **Custom GPTs and Gems** — List all custom assistants built on deprecated models (o3, GPT-4.5, or older Claude/Gemini versions)
- **API integrations** — Review scripts, plugins, or automation tools that call specific model versions
- **Saved prompts and templates** — Catalog prompt libraries, research briefs, and workflow templates optimized for legacy models
- **Project-level configurations** — Check ChatGPT Projects, Claude Projects, or Gemini Gems for hardcoded model references

# # # Map your research workflows to model functions

For each affected workflow, note what the model does:
- Handwriting transcription (pension files, probate records, cemetery registers)
- Document analysis and source extraction (land deeds, census records, wills)
- Research planning and hypothesis generation
- Citation formatting and bibliography generation
- Narrative writing and ancestor biography drafting

# # # Document current performance baselines

Before migrating, capture examples of current output quality:

- Save 3–5 sample prompts with their responses from legacy models
- Screenshot or export particularly successful analyses
- Note which model parameters you use (temperature, max tokens, system prompts)

# # Phase 2: Select Replacement Models

# # # Match tasks to model strengths (June 2026)

Based on recent benchmarks and genealogy-specific testing:

**For handwriting transcription and OCR:**
- **Primary:** Claude Opus 4.8 (4x better error detection, flags uncertain readings)
- **Secondary:** GPT-5.5 Instant (improved context memory for writer-specific patterns)
**For multi-document analysis and cross-referencing:**
- **Primary:** Claude Opus 4.8 (extended reasoning for probate packets, pension files)
- **Secondary:** Gemini 3.5 Flash (fast processing for batch operations)

**For research planning and structured output:**
- **Primary:** GPT-5.5 Instant (conversational memory, template refinement)
- **Secondary:** Claude Sonnet 4.5 (balance of speed and depth)

**For citation and bibliography generation:**
- **Primary:** GPT-5.5 Instant (follows Genealogical Proof Standard formatting reliably)
- **Secondary:** Claude Opus 4.8 (precise source parsing)

# # # Test replacement models with saved examples

Run your Phase 1 sample prompts through candidate replacement models:

1. Use identical prompts on new models
2. Compare output quality, accuracy, and formatting
3. Note any performance differences (speed, token usage, hallucination patterns)
4. Adjust prompts if necessary to match new model strengths

# # Phase 3: Execute the Migration

# # # For ChatGPT Projects and Custom GPTs

**Migrate Custom GPTs from o3 or GPT-4.5 to GPT-5.5 Instant:**

1. Open your Custom GPT in ChatGPT
2. Click "Configure" to access instructions, uploaded files, and capabilities
3. Copy all instructions, conversation starters, and uploaded knowledge files
4. Create a new Custom GPT using GPT-5.5 Instant as the base model
5. Paste instructions and re-upload knowledge files
6. Test with 5–10 representative prompts from your research history
7. Fine-tune instructions based on output differences (GPT-5.5 may need more explicit formatting instructions)
8. Update any shared links or bookmarks to point to the new Custom GPT

**Migrate ChatGPT Projects:**

1. Open the legacy Project
2. Navigate to Project Settings (three dots)
3. Copy all project-level instructions
4. Download all uploaded files (PDFs, GEDCOM files, spreadsheets)
5. Create a new Project with GPT-5.5 Instant as the default model
6. Paste project instructions and re-upload files
7. Review conversation history—if critical context exists, export key threads as markdown and re-upload as reference documents

# # # For Claude Projects (if migrating from older Claude models)

**Upgrade to Claude Opus 4.8 or Sonnet 4.6:**

1. Open your existing Claude Project
2. Click the model selector dropdown
3. Choose Claude Opus 4.8 for complex analysis or Sonnet 4.6 for faster iterations
4. Test with existing Project context—Claude Projects automatically maintain conversation history and uploaded files
5. No file re-upload required; existing knowledge carries forward

# # # For Google Gemini (migrating from Gemini 2.0 Flash models)

**Critical deadline: June 1, 2026:**

1. Update all API calls from `gemini-2.0-flash` or `gemini-2.0-flash-lite` to `gemini-3.5-flash` or `gemini-3.1-flash-lite`
2. For Gemini web interface users, Gems will automatically use the latest models
3. **If using the API:** Change model parameter in your code:

```python
# OLD (fails after June 1, 2026):
model="gemini-2.0-flash"

# NEW:
model="gemini-3.5-flash"
```

4. For OpenAI-compatible endpoints switching to Gemini:

```python
from openai import OpenAI
client = OpenAI(
api_key="your-gemini-api-key",
base_url="https://generativelanguage.googleapis.com/v1beta/openai/"
)
response = client.chat.completions.create(
model="gemini-3.5-flash", # Map from GPT models
messages=[{"role": "user", "content": "Your prompt"}]
)
```

# # # For Perplexity, Grok, and other platforms

Most platforms auto-update to latest models, but verify:

- Check account settings for model version locks
- Review any API integrations for hardcoded model names
- Update bookmarks if model-specific URLs exist

# # Phase 4: Migrate Custom Instructions and Memory

# # # Transfer ChatGPT custom instructions and memory to new models or platforms

**If staying within ChatGPT:**

1. Go to Settings → Personalization → Custom Instructions
2. Copy both "What would you like ChatGPT to know about you?" and "How would you like ChatGPT to respond?"
3. These automatically apply to GPT-5.5 Instant—no action needed if migrating within the same platform

**If migrating ChatGPT → Gemini:**

1. In ChatGPT: Settings → Data Controls → Export Data
2. Download your data archive and extract `chat.html`
3. In Gemini: Create a new Gem
4. Under "Instructions for Gemini," paste your custom instructions from ChatGPT
5. Upload `chat.html` to the Gem's Knowledge section
6. For memories, go to ChatGPT Settings → Personalization → Manage Memories
7. Review each memory—delete outdated or incorrect entries
8. Copy remaining memories one-by-one into Gemini's "Instructions" field (Gemini doesn't have a separate memory system; fold memories into instructions)

**If migrating ChatGPT → Claude:**

1. Export ChatGPT custom instructions and memories as above
2. Claude doesn't have a "memory" feature—incorporate key preferences into your Project instructions or a reference document
3. Upload chat exports as `.txt` or `.md` files to your Claude Project for context

# # # Clean up legacy memories before migration

Review for outdated information:

- Incorrect ancestor dates or places discovered during research
- Old surname spellings you've since corrected
- Deprecated file structures or naming conventions
- Removed research focuses (completed family lines)

# # Phase 5: Update Saved Prompts and Prompt Libraries

# # # Adjust prompts for new model capabilities

**For Claude Opus 4.8 (improved self-correction):**

Add explicit uncertainty requests to prompts:

- **Old prompt:** "Transcribe this 1880 probate document."
- **New prompt:** "Transcribe this 1880 probate document. Flag any words or dates where handwriting is unclear or you're uncertain of accuracy."

**For GPT-5.5 Instant (memory transparency):**
Reference prior research explicitly:

- **Old prompt:** "Analyze this land deed."
- **New prompt:** "Analyze this land deed. Reference my previous research on the Clark family in Creek Nation, and note which prior files or conversations inform your interpretation."

**For Gemini 3.5 Flash (speed optimization):**

Streamline batch prompts:

- **Old prompt:** "Transcribe this cemetery record, then analyze burial patterns, then suggest next research steps."
- **New prompt:** "Transcribe this cemetery record. Output: Name, birth date, death date, plot location." (Submit analysis as a separate, faster query)

# # # Create version-controlled prompt templates

Save prompts in markdown files with metadata:
```markdown
# Probate Extraction Template
**Model:** Claude Opus 4.8
**Updated:** June 1, 2026
**Use case:** Extract structured data from handwritten probate inventories

**Prompt:**
Analyze this probate inventory. Extract:
- Decedent name and death date
- Executor/administrator names
- All property descriptions with appraised values
- Names of heirs and their relationships
- Court and filing date

Flag any uncertain readings or date conflicts.
```

# # Phase 6: Test, Monitor, and Optimize

# # # Run parallel testing during transition period

For 2–4 weeks, run critical workflows on both old and new models:

1. Submit identical prompts to legacy and replacement models
2. Compare outputs for accuracy, completeness, and formatting
3. Document any regressions (tasks the new model performs worse)
4. Adjust prompts or switch models for specific tasks if needed

# # # Monitor for post-migration issues
After full migration:

- Track any increase in hallucinations or formatting errors
- Note if citation accuracy changes
- Monitor token usage and costs (new models may have different pricing)
- Check for changes in processing speed

# # # Build rollback procedures

Until legacy models fully retire:

- Keep access to Custom GPTs built on o3 or GPT-4.5 until August/June retirement dates
- Save `.bak` files of critical prompts and instructions
- Document what worked in legacy models vs. replacements

# # Phase 7: Update Documentation and Training

# # # Revise research guides and workflows

If you teach genealogy workshops or share methods:

1. Update slides, handouts, and blog posts with new model names
2. Replace screenshots showing deprecated models
3. Revise prompt examples to leverage new model features (e.g., Claude Opus 4.8's uncertainty flags)

# # # Train collaborators and research teams

For genealogy societies or research groups:

- Host a migration workshop showing before/after workflows
- Share updated prompt libraries with version notes
- Create troubleshooting guides for common migration issues

# # # Document your migration decisions

Create a log for future reference:

- Which models you chose for which tasks
- Prompt adjustments you made
- Performance differences you observed
- Cost or speed changes

# # Platform-Specific Migration Checklists

# # # ChatGPT (o3 and GPT-4.5 retirement)

- [ ] Export all Custom GPTs' instructions and files
- [ ] Create new Custom GPTs using GPT-5.5 Instant
- [ ] Test with representative prompts
- [ ] Update shared links
- [ ] Copy Project instructions and re-upload files
- [ ] Review and clean ChatGPT Memories
- [ ] Update API calls from `o3` or `gpt-4.5` to `gpt-5.5-instant`
- [ ] Test all integrations (Zapier, Make, custom scripts)
- [ ] Deadline: August 26, 2026 (o3) / June 27, 2026 (GPT-4.5)

# # # Google Gemini (2.0 Flash retirement)

- [ ] Update API calls from `gemini-2.0-flash` to `gemini-3.5-flash`
- [ ] Update API calls from `gemini-2.0-flash-lite` to `gemini-3.1-flash-lite`
- [ ] Test all scripts and integrations
- [ ] Verify Gems are using updated models (automatic for web users)
- [ ] Check for any hardcoded model names in automation tools
- [ ] **Deadline: June 1, 2026** (TODAY—migrations failing after this date)
# # # Anthropic Claude (if upgrading to Opus 4.8)

- [ ] Open existing Projects and switch model to Opus 4.8
- [ ] Test with existing context (files persist automatically)
- [ ] Adjust prompts to request uncertainty flags
- [ ] Update API calls to `claude-opus-4-8` or `claude-sonnet-4-6`
- [ ] No hard deadline (but Opus 4.8 offers significant quality improvements)

# # # Cross-platform migration (ChatGPT → Gemini or Claude)

- [ ] Export ChatGPT data (Settings → Export Data)
- [ ] Extract `chat.html` from archive
- [ ] Copy custom instructions and memories
- [ ] Create new Gem (Gemini) or Project (Claude)
- [ ] Upload chat history and knowledge files
- [ ] Test identity and context retention
- [ ] Update browser bookmarks and shortcuts
- [ ] Notify collaborators of platform change

# # Troubleshooting Common Migration Issues

**Issue:** New model gives different answers to identical prompts
**Solution:** Check if the new model requires more explicit instructions. GPT-5.5 may need formatting examples; Claude Opus 4.8 responds well to "flag uncertainties" requests.
**Issue:** Custom GPT or Gem no longer accesses uploaded files
**Solution:** Re-upload files after migration. File associations don't always transfer when switching base models.

**Issue:** API calls fail with "model not found" errors
**Solution:** Verify exact model name spelling. Use `gpt-5.5-instant`, `claude-opus-4-8`, or `gemini-3.5-flash` (not `GPT-5.5`, `Claude-Opus-4.8`, or `Gemini-3.5-Flash`—casing matters).

**Issue:** Increased costs after migration
**Solution:** Check new model pricing. Opus 4.8 costs $5/$25 per million tokens; GPT-5.5 Instant has different rates than o3. For budget-conscious workflows, use Sonnet 4.6 or Gemini 3.5 Flash instead.
**Issue:** Memories or context seem lost
**Solution:** For ChatGPT, go to Settings → Personalization → Manage Memories and verify they weren't auto-deleted. For cross-platform migrations, manually re-enter key preferences.

# # Post-Migration Best Practices

# # # Maintain a multi-model strategy

Don't rely on a single model:

- Use Claude Opus 4.8 for complex handwriting transcription
- Use GPT-5.5 Instant for research planning and narrative writing
- Use Gemini 3.5 Flash for batch operations and quick lookups

# # # Set up automated model retirement alerts

Create calendar reminders or use tools like Releasebot to monitor deprecation announcements from OpenAI, Anthropic, and Google. Future retirements will follow similar 3–6 month notice periods.

# # # Archive successful workflows

When you find a prompt-model combination that works well:

- Save it in a version-controlled document (markdown or text file)
- Include the model name, date tested, and example outputs
- Share with your research community or genealogy society

# # # Plan for continuous evolution

AI models update every 2–4 weeks in 2026. Build flexibility into your workflows:

- Avoid hardcoding model names where possible
- Use feature-based selection (e.g., "best handwriting model") rather than specific versions
- Test new models quarterly to stay current with improvements

***

**Summary:** Migration is a multi-phase process requiring audit, testing, and documentation. With o3 retiring in 8 weeks, GPT-4.5 in 3.5 weeks, and Gemini 2.0 Flash retired today, researchers should prioritize API updates and Custom GPT migrations immediately. Claude Opus 4.8's improved accuracy and GPT-5.5's memory transparency offer meaningful upgrades worth the transition effort.

The last 72 hours brought significant updates across multiple AI platforms (including model retirements that will impact your previous wo...

05/26/2026

The question came up yesterday when I was visiting with a family member. "Can a spouse of a veteran be buried in a National Cemetery, even if the veteran is not?"

Can a Veteran’s Spouse Get Burial Benefits?

Yes, a veteran’s spouse may be eligible for burial benefits provided certain criteria are met. The spouse of a veteran may be entitled to burial in a VA national cemetery alongside the veteran. This benefit extends to the spouse even if the veteran is buried in a private cemetery.
Requirements for Spouse Eligibility

For a veteran’s spouse to qualify for burial benefits, the marriage to the veteran must have been valid at the time of the veteran’s passing. Additionally, the spouse must not have been remarried after the veteran’s death.
Benefits Available to the Spouse

Upon the death of a veteran, the spouse may be eligible for a burial flag, a Presidential Memorial Certificate, and in some cases, financial assistance towards the burial or cremation costs.
Final Thoughts on Veteran Burial Benefits for Spouses

It is essential for veterans and their families to be aware of the burial benefits available to them. The VA provides a range of honors to commemorate the service of veterans, and these benefits extend to the spouses of eligible veterans. By understanding the eligibility criteria and benefits available, families can ensure that their loved ones receive the recognition and honors they deserve.

Send a message to learn more

05/18/2026

It'sNOT too late to save your virtual spot. https://us02web.zoom.us/meeting/register/xoVLBV6PSvKr9TfniceQxA

05/15/2026

Memorial Day is fast approaching. Even though it is officially the day to remember those who have made the ultimate sacrifice, many people use the occasion to visit and decorate the graves of their loved ones, regardless of their military status. Here's a good guide by William Dollarhide to keep handy for your visit to a cemetery as well as your cemetery research.

05/15/2026

Monitoring new digital collections for relevant items

Periodically feed AI announcement feeds or blog posts from archives and major genealogy platforms and ask it to extract only those new collections that match your focus surnames, localities, or time periods, creating a watch list with links and dates.

05/13/2026

Girl Scout Cookies documentary:

View the e-Edition for Wednesday, May 13, 2026

05/12/2026

Use this prompt In a product that uses Perplexity’s 19‑model Computer:

“Identify all major online record sets for Muskogee County, Oklahoma, 1880–2020, across Ancestry, FamilySearch, state archives, and local libraries. Since Muskogee County is in both the Creek and Cherokee nation reservations, be sure to identify those relevant records sets as well to include both contexts, OHS American Indian Archives, FamilySearch Cherokee/Dawes guides, and NARA Dawes/BIA references. Produce a table with collection name, date span, URL and Tribal/Jurisdiction Scope.”

05/09/2026

Genealogy Platforms Weekly Briefing: May 2–9, 2026

The past seven days (May 2–9, 2026) have been relatively quiet across major genealogy platforms, with no major feature announcements or tool launches reported. Activity continues to center on incremental record additions, particularly in newspaper databases, as platforms maintain the steady expansion patterns established in April 2026.

Platform Updates

FamilySearch.org
FamilySearch published its April 2026 monthly update on April 9, adding over 1 billion free historical records focused heavily on the United States (1 billion records including public/vital records, military/cemetery records, and immigration/naturalization), United Kingdom (13 million vital, military, and emigration records), and Ireland (10 million civil registrations and session court records). No new platform features or tools were reported this week. The platform continues its shift to monthly (rather than weekly) record updates, with the next monthly report expected in early June 2026.

Ancestry.com
No feature updates or major collection announcements were reported for the week of May 2–9. At RootsTech 2026 (March 5–7), Ancestry announced a $450 million investment in new content and showcased Revolutionary War pension files with 10 million searchable names via full-text search and handwriting recognition. These initiatives are rolling out gradually throughout 2026, but no specific updates tied to this past week were documented.

MyHeritage.com
MyHeritage's most recent collection update covered April 1–30, 2026, adding 6 new and 10 updated collections, including Austria Vienna Catholic Church Births and Baptisms 1850–1920, California Deaths 1905–1939, Hungary World War I Casualty Lists 1914–1919, and Netherlands Names Stories in Newspapers from OldNews.com. No updates specific to May 2–9 were reported. Earlier this year, MyHeritage announced its transition to Whole Genome Sequencing (WGS) for DNA tests and expansion by 6.6 billion historical records in 2025.

Newspapers.com
GenealogyBank (a competitor subscription newspaper site) announced 14 titles added or updated for May 2026 on May 4. Newspapers.com itself has not published a dedicated May update, but April/May aggregate reports indicate over 1,000 new titles added across various free and subscription newspaper databases. Earlier reports from January and March 2026 documented 218 and 90 new newspapers respectively, reinforcing the pattern of large monthly batches.

Elephind.com
Elephind continues operating with a refreshed interface and over 13 million newspaper pages indexed, with additional pages queued. No updates specific to the May 2–9 window were found.

Archive-It.org
The Internet Archive's Archive-It program published recent posts about community web-archiving initiatives, including projects digitizing local LGBTQ+ organizational records, photographs, and related materials through partner institutions. No genealogy-specific announcements for May 2–9 were identified.

AdvantageArchives.com
Advantage Archives Collections (735 collections from 44 U.S. states) continues expanding, with updates documented in April/May aggregate newspaper reports showing participation in the 1,000+ new titles added across databases. No platform-specific updates for this week were reported.

Notable Examples to Try This Week

1. Revolutionary War pension files on Ancestry: Search the newly full-text-searchable Revolutionary War pension files (10 million searchable names announced at RootsTech 2026) for collateral surnames, place names, and enslaved individuals mentioned in narrative sections. Even if you explored these files before, the handwriting recognition and full-text capability makes previously "invisible" mentions searchable for the first time.

2. FamilySearch April 2026 U.S. or Ireland collections: Run a targeted search in the freshly expanded April 2026 collections (1 billion U.S. records; 10 million Ireland civil registrations and session court records) for brick-wall ancestors in those jurisdictions. Even quiet collection expansions in civil registrations or parish material can break open stalled lines.

3. MyHeritage Netherlands OldNews.com newspaper collection: Explore the new Netherlands Names Stories in Newspapers from OldNews.com collection added in April 2026. Search for Dutch surnames or emigration-related mentions to surface family stories, departure notices, or community context.

Send a message to learn more

05/07/2026

Why obituaries are a sweet spot for AI help

- Newspaper and library guides call historical newspapers “one of the most important resources” for genealogists, with obituaries, marriage notices, and legal ads adding rich detail to the bare bones of civil records.
- How‑to articles stress that obituaries can supply names of relatives, residences, burial places, religious affiliation, occupations, and cause of death, but also warn that they must be checked against other sources for accuracy.
- Platforms and tutorials recommend systematic approaches—writing down known facts first, then using time, place, and variant spellings to target searches in databases like NewspaperArchive, Newspapers.com, Chronicling America, and library‑licensed collections.

AI fits best **after** you’ve located candidate obituaries: extracting people, relationships, places, and dates into structured notes, plus synthesizing them with census, vital, and probate data you already have.