Steve Little
@digitalarchivst.bsky.social
AI Program Director, National Genealogical Society; Family History AI Show podcast; Founder: AIGenealogyInsights.com
Husband, Dad, birder, chess dilettante, film & TV fan, AI & genetic genealogist, Methodist pastor, photog, reader, writer, skygazer, NC, VA
Husband, Dad, birder, chess dilettante, film & TV fan, AI & genetic genealogist, Methodist pastor, photog, reader, writer, skygazer, NC, VA
You will hear of launches and rumors of launches—but here's how to test Gemini 3 yourself. Pick one stubborn document or photo. Run it through our #genealogy research-tuned Gem today (Gemini 2.5). Save everything. When G3 lands, rerun the exact same prompts. If it's real, you'll feel the difference.
November 18, 2025 at 12:16 AM
You will hear of launches and rumors of launches—but here's how to test Gemini 3 yourself. Pick one stubborn document or photo. Run it through our #genealogy research-tuned Gem today (Gemini 2.5). Save everything. When G3 lands, rerun the exact same prompts. If it's real, you'll feel the difference.
Know Your Data. Know Your Model. Know Your Limits.
Prof Mollick articulates why this framework matters: "You need to know specifically what YOUR AI is good at, not what AIs are good at on average."
Test models on YOUR real genealogical work—not generic assumptions.
Link ➡️
Prof Mollick articulates why this framework matters: "You need to know specifically what YOUR AI is good at, not what AIs are good at on average."
Test models on YOUR real genealogical work—not generic assumptions.
Link ➡️
November 12, 2025 at 7:12 PM
Know Your Data. Know Your Model. Know Your Limits.
Prof Mollick articulates why this framework matters: "You need to know specifically what YOUR AI is good at, not what AIs are good at on average."
Test models on YOUR real genealogical work—not generic assumptions.
Link ➡️
Prof Mollick articulates why this framework matters: "You need to know specifically what YOUR AI is good at, not what AIs are good at on average."
Test models on YOUR real genealogical work—not generic assumptions.
Link ➡️
The 'AI slop' critics point to? That's just beginner output. Learning LLMs is like learning to ride a bike—you can't read your way to mastery. Smart people try once, produce dreck, then dismiss the tool. But they're confusing their inexperience with the tool's limits.
November 11, 2025 at 3:14 AM
The 'AI slop' critics point to? That's just beginner output. Learning LLMs is like learning to ride a bike—you can't read your way to mastery. Smart people try once, produce dreck, then dismiss the tool. But they're confusing their inexperience with the tool's limits.
A Sunday morning reflection:
“Every design choice expresses a vision of humanity.” —Pope Leo XIV
There are better and worse ways to steer the AI revolution. The worst hands our values to a few unaccountable corporate tech boards while governments dither or sell-out. (1/4)
“Every design choice expresses a vision of humanity.” —Pope Leo XIV
There are better and worse ways to steer the AI revolution. The worst hands our values to a few unaccountable corporate tech boards while governments dither or sell-out. (1/4)
November 9, 2025 at 11:44 AM
A Sunday morning reflection:
“Every design choice expresses a vision of humanity.” —Pope Leo XIV
There are better and worse ways to steer the AI revolution. The worst hands our values to a few unaccountable corporate tech boards while governments dither or sell-out. (1/4)
“Every design choice expresses a vision of humanity.” —Pope Leo XIV
There are better and worse ways to steer the AI revolution. The worst hands our values to a few unaccountable corporate tech boards while governments dither or sell-out. (1/4)
This spring, new AI photo tools arrived. Genealogists experimented. Then quickly we realized: these tools don't "restore" historical photos—they may fabricate new faces.
By August, these "restorations" were everywhere, unlabeled, corrupting our shared historical record. #genealogy
By August, these "restorations" were everywhere, unlabeled, corrupting our shared historical record. #genealogy
November 7, 2025 at 3:39 PM
This spring, new AI photo tools arrived. Genealogists experimented. Then quickly we realized: these tools don't "restore" historical photos—they may fabricate new faces.
By August, these "restorations" were everywhere, unlabeled, corrupting our shared historical record. #genealogy
By August, these "restorations" were everywhere, unlabeled, corrupting our shared historical record. #genealogy
The Coalition for Responsible AI in #Genealogy has released its new official position statement: "Protecting Trust in Historical Images."
While AI "photo restoration" tools can be helpful, they also risk altering or fabricating features, misleading researchers. Link and more ►
While AI "photo restoration" tools can be helpful, they also risk altering or fabricating features, misleading researchers. Link and more ►
November 7, 2025 at 2:33 AM
The Coalition for Responsible AI in #Genealogy has released its new official position statement: "Protecting Trust in Historical Images."
While AI "photo restoration" tools can be helpful, they also risk altering or fabricating features, misleading researchers. Link and more ►
While AI "photo restoration" tools can be helpful, they also risk altering or fabricating features, misleading researchers. Link and more ►
Hello friends and fellow researchers! Episode 36 of the podcast, “A Simple Path to Better Prompts,” is out this morning, and I’d like to give you a long walk-through of the Tip of the Week, "Building Reusable Research Prompts," which I shared in this episode. (Link in next comment.) #genealogy
November 3, 2025 at 11:38 PM
Hello friends and fellow researchers! Episode 36 of the podcast, “A Simple Path to Better Prompts,” is out this morning, and I’d like to give you a long walk-through of the Tip of the Week, "Building Reusable Research Prompts," which I shared in this episode. (Link in next comment.) #genealogy
I've written thousands of prompts over the last three years; here are the sixteen that have stood the test of time. These are my best that I still use every day, gathered in one quick copy-and-paste guide. I keep these at my fingertips in a clipboard manager, and reuse them all day long.
October 31, 2025 at 10:03 AM
I've written thousands of prompts over the last three years; here are the sixteen that have stood the test of time. These are my best that I still use every day, gathered in one quick copy-and-paste guide. I keep these at my fingertips in a clipboard manager, and reuse them all day long.
This isn't a future problem—people are already experiencing the impact. AI should be made to pay its own way. Data centers should pay higher tax rates. "Taxing compute" needs to be a part of the common vocabulary. We need governments worthy of this challenge.
October 23, 2025 at 4:43 PM
This isn't a future problem—people are already experiencing the impact. AI should be made to pay its own way. Data centers should pay higher tax rates. "Taxing compute" needs to be a part of the common vocabulary. We need governments worthy of this challenge.
Some folks think AI is a bubble. That is wishful thinking. Companies will fail, but the AI revolution will roll-on. The difference is the rate of change. Even if progress stopped today, it will take decades to discover the capabilities already under our fingertips.
October 23, 2025 at 4:43 PM
Some folks think AI is a bubble. That is wishful thinking. Companies will fail, but the AI revolution will roll-on. The difference is the rate of change. Even if progress stopped today, it will take decades to discover the capabilities already under our fingertips.
Asking a model to critique or improve your prompts--called "meta-prompting"--is a strong and good use of these tools. And both OpenAI and Anthropic has tools to help you craft better prompts.
console.anthropic.com/dashboard
console.anthropic.com/dashboard
September 29, 2025 at 9:41 PM
Asking a model to critique or improve your prompts--called "meta-prompting"--is a strong and good use of these tools. And both OpenAI and Anthropic has tools to help you craft better prompts.
console.anthropic.com/dashboard
console.anthropic.com/dashboard
The barrier between an idea and a compelling image has essentially disappeared.
September 14, 2025 at 2:35 AM
The barrier between an idea and a compelling image has essentially disappeared.
Here's a prompt to add a spectral wire sculpture to an image, inspired by the context and contents of the image (link to full prompt in first comment).
September 12, 2025 at 1:07 PM
Here's a prompt to add a spectral wire sculpture to an image, inspired by the context and contents of the image (link to full prompt in first comment).
The prompt to generate the image was itself crafted with a strong writer, o3, and is interesting to read in its own.
I crafted the prompt below with an earlier version of OpenAI's o3 model. The image was generated with their GPT-4o model. (Full prompt in image.)
I crafted the prompt below with an earlier version of OpenAI's o3 model. The image was generated with their GPT-4o model. (Full prompt in image.)
June 14, 2025 at 5:11 AM
The prompt to generate the image was itself crafted with a strong writer, o3, and is interesting to read in its own.
I crafted the prompt below with an earlier version of OpenAI's o3 model. The image was generated with their GPT-4o model. (Full prompt in image.)
I crafted the prompt below with an earlier version of OpenAI's o3 model. The image was generated with their GPT-4o model. (Full prompt in image.)
Virginia’s Great Seal (1776)—devised by revolutionaries George Wythe and George Mason—shows the Roman goddess Virtus, spear grounded, sword poised, standing over a deposed, crown-fallen tyrant; its motto “Sic semper tyrannis” (“Thus always to tyrants”) proclaims unending resistance to oppression.
June 14, 2025 at 5:11 AM
Virginia’s Great Seal (1776)—devised by revolutionaries George Wythe and George Mason—shows the Roman goddess Virtus, spear grounded, sword poised, standing over a deposed, crown-fallen tyrant; its motto “Sic semper tyrannis” (“Thus always to tyrants”) proclaims unending resistance to oppression.
Predicted, now here: GPT-4o generates accurate in-image text. However, "AI Reconstruction" by GPT-4o prompts debate: what do we call these things?!?
Image: The man on the left is my grandfather; the man on the right is not.
aigenealogyinsights.com/2025/06/10/e...
Image: The man on the left is my grandfather; the man on the right is not.
aigenealogyinsights.com/2025/06/10/e...
June 11, 2025 at 4:11 AM
Predicted, now here: GPT-4o generates accurate in-image text. However, "AI Reconstruction" by GPT-4o prompts debate: what do we call these things?!?
Image: The man on the left is my grandfather; the man on the right is not.
aigenealogyinsights.com/2025/06/10/e...
Image: The man on the left is my grandfather; the man on the right is not.
aigenealogyinsights.com/2025/06/10/e...
Ep.25 of Family History AI Show: GPT-4o’s photorealism & OpenAI’s o3 geolocation reasoning impress—accurate within blocks. Plus, Google Gemini Veo 3 hints at next-gen video/audio generation.
June 11, 2025 at 4:11 AM
Ep.25 of Family History AI Show: GPT-4o’s photorealism & OpenAI’s o3 geolocation reasoning impress—accurate within blocks. Plus, Google Gemini Veo 3 hints at next-gen video/audio generation.
Some models are better with words than others. I was using OpenAI's o3 to help me understand some text, so I prompted a summary, and it responded:
"ignorance is the seedbed of unintended harm"
"ignorance is the seedbed of unintended harm"
June 5, 2025 at 1:56 AM
Some models are better with words than others. I was using OpenAI's o3 to help me understand some text, so I prompted a summary, and it responded:
"ignorance is the seedbed of unintended harm"
"ignorance is the seedbed of unintended harm"
Ask Hard Questions! Drop your toughest AI-in-genealogy questions below. Ahead of Sunday’s NGS panel on Guiding AI Principles—Accuracy, Disclosure, Privacy, Education, Compliance—I’ll be on stage with Lynn Broderick and BCG president David Ouimette. What should we tackle? ⬇️
May 23, 2025 at 4:13 PM
Ask Hard Questions! Drop your toughest AI-in-genealogy questions below. Ahead of Sunday’s NGS panel on Guiding AI Principles—Accuracy, Disclosure, Privacy, Education, Compliance—I’ll be on stage with Lynn Broderick and BCG president David Ouimette. What should we tackle? ⬇️
o3 make geolocation trivial.
chatgpt.com/share/6825e9...
chatgpt.com/share/6825e9...
May 15, 2025 at 1:19 PM
o3 make geolocation trivial.
chatgpt.com/share/6825e9...
chatgpt.com/share/6825e9...
My improv on "Please" and "Thank you": the wink.😉
April 22, 2025 at 11:36 PM
My improv on "Please" and "Thank you": the wink.😉
Meet Linka, an AI assistant introduced by Borland Genetics, who helps Kevin Borland code the site to help users. If you'd like to learn how beginners can create your own AI assistants, Andrew Redfern is teaching a free webinar on saved prompts. #genealogy
Details on both in comments. 🔗⬇️
Details on both in comments. 🔗⬇️
April 3, 2025 at 12:59 AM
Meet Linka, an AI assistant introduced by Borland Genetics, who helps Kevin Borland code the site to help users. If you'd like to learn how beginners can create your own AI assistants, Andrew Redfern is teaching a free webinar on saved prompts. #genealogy
Details on both in comments. 🔗⬇️
Details on both in comments. 🔗⬇️
For I4GG AI Day, please enjoy my "AI Resources for Genealogists" infographic. Mark @makingfamilyhistory.com Thompson and I are excited to be collaborating with CeCe Moore and the Institute for Genetic Genealogy to present I4GG AI Day 2025 on Fri 28 Mar 2025, both in-person and virtually. #genealogy
March 25, 2025 at 8:16 PM
For I4GG AI Day, please enjoy my "AI Resources for Genealogists" infographic. Mark @makingfamilyhistory.com Thompson and I are excited to be collaborating with CeCe Moore and the Institute for Genetic Genealogy to present I4GG AI Day 2025 on Fri 28 Mar 2025, both in-person and virtually. #genealogy
Fun Prompt Friday! Prompt AI to sketch your genealogical roadmap: detailed locality guides full of the places, people, and resources shaping your family history. #genealogy #familyHistory #localityGuides
Link below 🔗👇
Link below 🔗👇
March 1, 2025 at 12:14 AM
Fun Prompt Friday! Prompt AI to sketch your genealogical roadmap: detailed locality guides full of the places, people, and resources shaping your family history. #genealogy #familyHistory #localityGuides
Link below 🔗👇
Link below 🔗👇
Four major companies released AI research agents (Google, DeepSeek, Perplexity, xAI, OpenAI), but OpenAI's Deep Research stands apart in comprehensiveness, despite high costs ($20/mo: 10 queries, $200/mo: 120 queries). See for yourself!
Link below 🔗👇
Link below 🔗👇
February 27, 2025 at 6:24 PM
Four major companies released AI research agents (Google, DeepSeek, Perplexity, xAI, OpenAI), but OpenAI's Deep Research stands apart in comprehensiveness, despite high costs ($20/mo: 10 queries, $200/mo: 120 queries). See for yourself!
Link below 🔗👇
Link below 🔗👇