Arnab Nandi
@arnab.org
https://arnab.org | human-in-the-loop data infrastructure | Professor, Computer Science at Ohio State | Co-founder, @thesteamfactory, @hackohio, @mobikitinc (acq.)
Where is the Cursor for SQL?
CoQuery is a minimal vibe-coded SQL playground powered by SQLite WASM and Google Gemini language models. There is no backend, everything is in the webpage frontend.
github.com/arnabdotorg/...
CoQuery is a minimal vibe-coded SQL playground powered by SQLite WASM and Google Gemini language models. There is no backend, everything is in the webpage frontend.
github.com/arnabdotorg/...
August 26, 2025 at 9:41 PM
Where is the Cursor for SQL?
CoQuery is a minimal vibe-coded SQL playground powered by SQLite WASM and Google Gemini language models. There is no backend, everything is in the webpage frontend.
github.com/arnabdotorg/...
CoQuery is a minimal vibe-coded SQL playground powered by SQLite WASM and Google Gemini language models. There is no backend, everything is in the webpage frontend.
github.com/arnabdotorg/...
I decided to vibe code a quick "CSE 3241 Game Show" using Claude Artifacts and Gemini to help students study for their finals. I was very pleasantly surprised by how easy the process is. Figured I'd share the process, and some design principles I used! www.linkedin.com/pulse/vibe-c...
April 4, 2025 at 5:19 PM
I decided to vibe code a quick "CSE 3241 Game Show" using Claude Artifacts and Gemini to help students study for their finals. I was very pleasantly surprised by how easy the process is. Figured I'd share the process, and some design principles I used! www.linkedin.com/pulse/vibe-c...
This came up during a class discussion. Apple **has** launched a dedicated social network before. In 2010, iTunes Ping was launched to build out a music-focused social network: en.wikipedia.org/wiki/ITunes_...
February 19, 2025 at 4:45 AM
This came up during a class discussion. Apple **has** launched a dedicated social network before. In 2010, iTunes Ping was launched to build out a music-focused social network: en.wikipedia.org/wiki/ITunes_...
Love it. I'm just thinking, the obvious product name would be Excel Multiverse
February 5, 2025 at 6:06 AM
Love it. I'm just thinking, the obvious product name would be Excel Multiverse
The current excitement about research agents (@emollick.bsky.social has great notes) reminded me of the portfolio of tools from Corpernic in the early 2000s, including a summarizer, a search aggregator, and a shopper.
Obviously the capabilities are wildly different, but the need has been around.
Obviously the capabilities are wildly different, but the need has been around.
February 4, 2025 at 3:51 PM
The current excitement about research agents (@emollick.bsky.social has great notes) reminded me of the portfolio of tools from Corpernic in the early 2000s, including a summarizer, a search aggregator, and a shopper.
Obviously the capabilities are wildly different, but the need has been around.
Obviously the capabilities are wildly different, but the need has been around.
🔁@csail.mit.edu: Early Video of Jeff Bezos walking through the first office of Amazon.com Inc., featuring the "Programming Perl" camel book on a desk (Amazon was one of the biggest users of Perl / Mason for content generation.)
January 21, 2025 at 5:43 PM
🔁@csail.mit.edu: Early Video of Jeff Bezos walking through the first office of Amazon.com Inc., featuring the "Programming Perl" camel book on a desk (Amazon was one of the biggest users of Perl / Mason for content generation.)
OpenAI Sora notes.
- Performance: getting 5-10second videos within the minute of submitting, impressive.
- Quality: hard to iterate edits. They have remix and recut using existing video as seed, but it's at prompt level
- Composing: They have a composer mode, but there is a lot more that can be done
- Performance: getting 5-10second videos within the minute of submitting, impressive.
- Quality: hard to iterate edits. They have remix and recut using existing video as seed, but it's at prompt level
- Composing: They have a composer mode, but there is a lot more that can be done
January 9, 2025 at 5:51 PM
OpenAI Sora notes.
- Performance: getting 5-10second videos within the minute of submitting, impressive.
- Quality: hard to iterate edits. They have remix and recut using existing video as seed, but it's at prompt level
- Composing: They have a composer mode, but there is a lot more that can be done
- Performance: getting 5-10second videos within the minute of submitting, impressive.
- Quality: hard to iterate edits. They have remix and recut using existing video as seed, but it's at prompt level
- Composing: They have a composer mode, but there is a lot more that can be done
"LOTUS (LLMs Over Tables of Unstructured and Structured Data) provides a declarative programming model and an optimized query engine for serving reasoning-based query pipelines over structured & unstructured data" via a Pandas-like API: github.com/guestrin-lab... www.arxiv.org/abs/2407.11418
December 19, 2024 at 3:20 PM
"LOTUS (LLMs Over Tables of Unstructured and Structured Data) provides a declarative programming model and an optimized query engine for serving reasoning-based query pipelines over structured & unstructured data" via a Pandas-like API: github.com/guestrin-lab... www.arxiv.org/abs/2407.11418
December 19, 2024 at 3:03 PM
@joshpigford.com loving this. I'd love to know how many people are pressing the pre-generated prompt buttons, and how many are actual new questions.
This also reminds me of Office Contextual Ribbons + Mac Searchable Help -- LLMs now have the power to turbocharge them 1000x.
This also reminds me of Office Contextual Ribbons + Mac Searchable Help -- LLMs now have the power to turbocharge them 1000x.
December 17, 2024 at 9:18 PM
@joshpigford.com loving this. I'd love to know how many people are pressing the pre-generated prompt buttons, and how many are actual new questions.
This also reminds me of Office Contextual Ribbons + Mac Searchable Help -- LLMs now have the power to turbocharge them 1000x.
This also reminds me of Office Contextual Ribbons + Mac Searchable Help -- LLMs now have the power to turbocharge them 1000x.
This definitely helped my UX on here: Settings > Font Size > Larger
November 29, 2024 at 8:28 PM
This definitely helped my UX on here: Settings > Font Size > Larger
As "60fps" (ie, 16ms roundtrip latency) becomes reasonable, we have new challenges. UIs issue interactive queries in bursts -- Q2 may be issued BEFORE Q1 returns. Q3 can be blocked on Q1, but is part of "interactive" UX.
Here, we need to measure not just latency but Latency Constraint Violations:
Here, we need to measure not just latency but Latency Constraint Violations:
November 29, 2024 at 4:11 PM
As "60fps" (ie, 16ms roundtrip latency) becomes reasonable, we have new challenges. UIs issue interactive queries in bursts -- Q2 may be issued BEFORE Q1 returns. Q3 can be blocked on Q1, but is part of "interactive" UX.
Here, we need to measure not just latency but Latency Constraint Violations:
Here, we need to measure not just latency but Latency Constraint Violations:
My latest party trick is to make football analogy explainers out of PhD dissertations
November 15, 2024 at 11:34 PM
My latest party trick is to make football analogy explainers out of PhD dissertations
Sky Follower Bridge has been working really well for me
November 15, 2024 at 6:09 PM
Sky Follower Bridge has been working really well for me
Found this gem on @sigmodrecord: The very *FIRST* issue of FDT, brought to you by ACM SICFIDET(ACM Special Interest Committee on File Description & Translation).
Published in 1969, this would go on to grow into ACM @sigmod.
https://sigmodrecord.org/publications/sigmodRecord/6908/pdfs/984409.fm.pdf
Published in 1969, this would go on to grow into ACM @sigmod.
https://sigmodrecord.org/publications/sigmodRecord/6908/pdfs/984409.fm.pdf
June 7, 2023 at 5:17 PM
Found this gem on @sigmodrecord: The very *FIRST* issue of FDT, brought to you by ACM SICFIDET(ACM Special Interest Committee on File Description & Translation).
Published in 1969, this would go on to grow into ACM @sigmod.
https://sigmodrecord.org/publications/sigmodRecord/6908/pdfs/984409.fm.pdf
Published in 1969, this would go on to grow into ACM @sigmod.
https://sigmodrecord.org/publications/sigmodRecord/6908/pdfs/984409.fm.pdf
Why waste time on calls? Introducing PersonaGPT for FaceTime.
June 6, 2023 at 3:46 PM
Why waste time on calls? Introducing PersonaGPT for FaceTime.
From today’s White House announcement, US Department of Education:
“Artificial Intelligence and the Future of Teaching and Learning - Insights and Recommendations“
https://www2.ed.gov/documents/ai-report/ai-report.pdf
“Artificial Intelligence and the Future of Teaching and Learning - Insights and Recommendations“
https://www2.ed.gov/documents/ai-report/ai-report.pdf
May 24, 2023 at 4:06 AM
From today’s White House announcement, US Department of Education:
“Artificial Intelligence and the Future of Teaching and Learning - Insights and Recommendations“
https://www2.ed.gov/documents/ai-report/ai-report.pdf
“Artificial Intelligence and the Future of Teaching and Learning - Insights and Recommendations“
https://www2.ed.gov/documents/ai-report/ai-report.pdf
May 18, 2023 at 4:19 AM
Shout out to the interactive analytics developers at ATN who just got back from a stressful weekend
May 15, 2023 at 3:31 PM
Shout out to the interactive analytics developers at ATN who just got back from a stressful weekend
Gmail's new Generative AI "Help Me Write" vs ChatGPT, attempting to respond to an Area Chair invitation for ICDE 2024.
ChatGPT is still way better. This is something I expected Google to be much better at, given that it has 1B+ users.
(And to clarify, I did accept the invite!)
ChatGPT is still way better. This is something I expected Google to be much better at, given that it has 1B+ users.
(And to clarify, I did accept the invite!)
May 11, 2023 at 3:20 PM
Gmail's new Generative AI "Help Me Write" vs ChatGPT, attempting to respond to an Area Chair invitation for ICDE 2024.
ChatGPT is still way better. This is something I expected Google to be much better at, given that it has 1B+ users.
(And to clarify, I did accept the invite!)
ChatGPT is still way better. This is something I expected Google to be much better at, given that it has 1B+ users.
(And to clarify, I did accept the invite!)