iiinova — where your digital self keeps evolving

Peter Bull @peter.drivendata.org · 10/07/2026

Check out this challenge to understand real tutoring transcripts!

Reposted by Peter Bull

DrivenData @drivendata.org · 26/06/2026

A lot of “data usability” comes down to pretty unglamorous work: cleaning, standardizing, reconciling edge cases - but that’s the part that makes everything else possible. We wrote our approach here: drivendata.co/blog/last-mi...

drivendata.co

Solving the last-mile public data problem

Using "baked" data to transform public data repositories into analysis-ready resources

031

Peter Bull @peter.drivendata.org · 24/06/2026

First benchmark shared on K-12 AI Infrastructure Platform! Frontier models* don't outperform classical methods on grading science short answer questions. platform.k12-ai-infrastructure.org/benchmarks/2... (zero-short, few-shot, reasoning off).

platform.k12-ai-infrastructure.org

Competition - SAGE: Science Answer Grading & Evaluation

Compare models to automatically grade written responses to science questions

020

Peter Bull @peter.drivendata.org · 16/06/2026

Launching today: The K-12 AI Infrastructure Platform! platform.k12-ai-infrastructure.org Making AI better for students and educators. Learn more about the launch and our plans on the blog post here: drivendata.co/blog/k12-ai-...

010

Peter Bull @peter.drivendata.org · 01/04/2026

Excited to be speaking at Good Tech Summit in DC April 7 www.goodtechtogether.org/summit We’ll share a program focused on K-12 education and talk about investing in the foundations of AI: data, models, and benchmarks. We'll explore how these shape AI development in a field. Join us!

021

Peter Bull @peter.drivendata.org · 04/02/2026

🎉 Excited to launch this challenge! 🎉 Over a year of data collection, curation, and annotation that we undertook to produce a first-of-its-kind dataset. Help us build speech models that understand 2-5 year olds. $120k in prizes and huge impact! kidsasr.drivendata.org

000

Peter Bull @peter.drivendata.org · 13/10/2025

🚀 New release: cloudpathlib v0.23.0 🥧 Now with Python 3.14 (π) support! 📁 New copy & move methods mean you can reduce usage of shutil 🎉 Check out the full release and docs here: 👉 cloudpathlib.drivendata.org/stable/

000

Peter Bull @peter.drivendata.org · 10/10/2025

Super interesting work on new proposed columnar data file format called F3 with embedded wasm binary to decode the data 🤯 (which obviates the need for 3rd party library support). Favorable comparisons on compression, throughput and random reads to existing formats. db.cs.cmu.edu/papers/2025/...

000

Peter Bull @peter.drivendata.org · 08/10/2025

Very cool to see Wikimedia embracing LLM tools and launching a hybrid similarity search API and open source embeddings for Wikipedia! Also supports Q&A style queries. www.wikidata.org/wiki/Wikidat...

000

Peter Bull @peter.drivendata.org · 06/10/2025

Interesting to see empirical research coming out for LLMs as education aids. In this study, active use of LLMs helped CS students debug compiler errors. Removing LLM access demonstrated no lasting learning benefit from having had access to it... learninganalytics.upenn.edu/ryanbaker/IC...

000

Reposted by Peter Bull

Sara Beery @sarameghanbeery.bsky.social · 11/09/2025

Are you interested in #AIforConservation #AIforBiodiversity #AIforWildlife or #AIforNature?? Are you located in the Boston Area? If so, come join us!! The AI for Conservation Slack community is doing our first local-area Boston meetup, partnering with iNaturalist and TEDx Boston!

AI for Conservation Boston Meetup
Join us for an iNat bioblitz!!!
September 27th from 9am-12pm
Meet at umass Boston Quad at 9
Register here https://www.eventbrite.com/e/umass-boston-bioblitz-tickets-1626791971579?aff=oddtdtcreator

185

Peter Bull @peter.drivendata.org · 22/09/2025

We just shipped two major features for cloudpathlib ✨📦 ✨ ! First, http support—treat an URL like any other path (open, read_text, join). Second, compatibility with open and os Python built-ins for seamless transition of legacy code and third-party library support. cloudpathlib.drivendata.org

000

Peter Bull @peter.drivendata.org · 19/09/2025

Great opportunity to work on AI in conservation and biodiversity with Roland Kays! In-person in NC, check it out now since it is only open for a week: www.governmentjobs.com/careers/%7B0...

governmentjobs.com

Job Bulletin

State of North Carolina

000

Peter Bull @peter.drivendata.org · 20/08/2025

Exemplary FAQ for "Your Brain on ChatGPT: Accumulation of Cognitive Debt" www.brainonllm.com/faq I'd love to see more authors who are explicit about what NOT to claim based on a study, including wording for lay audiences that is not appropriate.

brainonllm.com

Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task

000

Peter Bull @peter.drivendata.org · 13/08/2025

Thought I would spot check a application someone was posting about 100% vibecoding. Can you spot the issue? Kudos to the LLM, this is verbatim from the fastapi docs. Sometimes verbatim from the docs is not what you want for your application though....

000

Peter Bull @peter.drivendata.org · 13/08/2025

Interesting announcement on a product from Astral! Similar model to one of the core @anacondainc.bsky.social lines of business.

010

Peter Bull @peter.drivendata.org · 08/08/2025

Enthusiastic to build on this generation of earth observation foundation embeddings like DeepMind's AlphaEarth (and more)! We already see some promising crop type (cereals vs. orchards) results and are exploring other use cases in climate resilience. deepmind.google/discover/blo...

000

Peter Bull @peter.drivendata.org · 01/08/2025

Very cool to see that marimo supports our cloudpathlib library for their file browser UI! Browse your S3, GCS, Azure buckets from your notebooks! docs.marimo.io/api/inputs/f...

docs.marimo.io

File Browser - marimo

The next generation of Python notebooks

000

Peter Bull @peter.drivendata.org · 25/07/2025

✨ 📦 ✨ Just released new Cookiecutter Data Science version with support for pixi and poetry as environment managers! Some of our top requested features ever. Upgrade and check it out now. cookiecutter-data-science.drivendata.org

010

Peter Bull @peter.drivendata.org · 18/07/2025

Now getting organic inbound for www.zambacloud.com, our wildlife imagery processing platform, from ChatGPT! 😲

000

Peter Bull @peter.drivendata.org · 16/07/2025

Just in case you thought speech-to-text worked for children, the third column is what Whisper does. Somehow in the third example it accesses my inner monologue... I guess that's why we're excited about our upcoming challenge! kidsasr.drivendata.org

010

Peter Bull @peter.drivendata.org · 14/07/2025

How are people managing code review for their AI coding agents? I do a first glance and it is obviously bad (e.g., didn't refactor repeated code), and now I've got half a dozen AI diffs for things that aren't good enough cluttering up my todo list with things to respond to....

000

Peter Bull @peter.drivendata.org · 09/07/2025

The best shortcut to how many experienced software engineers feel about AI is listening to the Primeagen's takes. Balanced perspectives on what's actually new, determinism, security, system complexity, what's promising, and what's not www.youtube.com/watch?v=vDWa...

000

Peter Bull @peter.drivendata.org · 07/07/2025

"Damn ChatGPT" your new summer jam about using ChatGPT as a therapist open.spotify.com/track/4umq06... (edited)

open.spotify.com

Maldito ChatGPT

Camilo · Maldito ChatGPT · Song · 2025

000

Peter Bull @peter.drivendata.org · 04/07/2025

Great article on the challenges of only surfacing the right info to LLMs and editing down what is not needed. If you've used a coding copilot or agent, you've seen this first hand many times. Output iterations are often polluted with code that came before. www.dbreunig.com/2025/06/22/h...

dbreunig.com

How Long Contexts Fail

Taking care of your context is the key to building successful agents. Just because there’s a 1 million token context window doesn’t mean you should fill it.

000

Peter Bull @peter.drivendata.org · 30/06/2025

BioCLIP2 looks like a stellar improvement! I'm excited to think about integrating into Zamba to for open-ended classification tasks run at scale on camera trap imagery. Definitely the potential to dramatically improve CT image utility. imageomics.github.io/bioclip-2/

000

Peter Bull @peter.drivendata.org · 25/06/2025

We've built so many low-fidelity prototypes in our HCD work. IMO vibecoding changes the feel of those prototypes, but doesn't change the process. Ask any designer—they'll tell you high-fidelity first iterations are often more distracting to clients than helpful. www.semafor.com/article/06/0...

000

Peter Bull @peter.drivendata.org · 23/06/2025

Check out this LLM circuit trace LLM for the text: '"The statement 'this statement is false' is." It goes through a logical contradictions node, but still outputs either "true" or "false" with the highest probabilities... www.anthropic.com/research/ope...

010

Peter Bull @peter.drivendata.org · 20/06/2025

A new preprint shows anonymization techniques for voices make transcription accuracy substantially worse for children versus adults. This is going to be a big challenge as we work on ASR for educational settings where we emphatically need both privacy and accuracy. arxiv.org/pdf/2506.00100

000

Peter Bull @peter.drivendata.org · 13/06/2025

😍 Incredible data storytelling about the power of conversation and human connection. Worth a read for good vibes! Based on the CANDOR corpus that we worked on. pudding.cool/2025/06/hell...

pudding.cool

30 minutes with a stranger

Watch hundreds of strangers talk for 30 minutes, and track how their moods change

000

Peter Bull @peter.drivendata.org · 06/06/2025

The gap between LLM prototype and production strikes again... in the worst possible place. www.propublica.org/article/trum...

000

Reposted by Peter Bull

European Space Agency @esa.int · 06/06/2025

📷 This week's @esaearth.esa.int #EarthFromSpace is a #Copernicus Sentinel-3 visible image of a thick plume of orange dust from the Sahara Desert over approximately 150 000 sq km of the eastern Atlantic Ocean on 7 May 2025🧪🌍 www.esa.int/ESA_Multimed...

The Sentinel-3 optical image shows a dense, orange plume of Saharan sand over approximately 150 000 sq km of the eastern Atlantic Ocean. The small islands of Cabo Verde peek out from beneath the clouds in the top left corner.

730141

Peter Bull @peter.drivendata.org · 30/05/2025

Very cool to see the multimodal conversation CANDOR dataset that we worked on used for a new paper on conversational agents! Gets agent feedback/training loops closer to the 7-38-55 rule than text only arxiv.org/abs/2505.15922

010

Reposted by Peter Bull

DrivenData @drivendata.org · 29/05/2025

You don't need to be a coder to use AI for wildlife research! 💻➡️🚫 With Zamba Cloud's new image support, simply upload photos, get species IDs, and even train custom models—all without writing a single line of code. www.zambacloud.com #WildlifeResearch

zambacloud.com

Zamba Cloud

012

Reposted by Peter Bull

Coral City Camera @coralcitycamera.bsky.social · 29/05/2025

A nurse shark takes a leisurely stroll through Coral City against a backdrop of thriving staghorn, elkhorn, brain, finger, and star corals #nurseshark #sharksofcoralcity #shark #leisurelystroll #elkhorn #staghorn #braincoral #coral #coralcitycamera #miami #portmiami #biscaynebay #coralcity

414783498

Peter Bull @peter.drivendata.org · 29/05/2025

What about the fact-checked news content?

000

Peter Bull @peter.drivendata.org · 28/05/2025

Foundation models for geospatial aren't there yet. This piece argues that unlike with language the information density of labeled data, and pretraining tasks aren't relevant enough. Worth a read if you do any geo AI christopherren.substack.com/p/geospatial...

christopherren.substack.com

Geospatial Foundational Disappointments

TLDR; After 10²¹ FLOPs and 500 B patches, IBM’s TerraMind beats a supervised U‑Net by just +2 mIoU on PANGAEA; losing on 5/9 tasks, most other GFMs do worse.

000

Reposted by Peter Bull

Stella Biderman @stellaathena.bsky.social · 23/05/2025

People keep plugging AI "Co-Scientists," so what happens when you ask them to do an important task like finding errors in papers? We built SPOT, a dataset of STEM manuscripts across 10 fields annotated with real errors to find out. (tl;dr not even close to usable) #NLProc arxiv.org/abs/2505.11855

411931

Peter Bull @peter.drivendata.org · 23/05/2025

This is a huge announcement for privacy-focused developers that don't want to do API calls, but the caniuse.com for built-in LLMs across browsers is going to be a shitshow. www.theverge.com/news/669528/...

theverge.com

Microsoft is opening its on-device AI models up to web apps in Edge

Edge continues to compete with Chrome.

000

Reposted by Peter Bull

Simon Willison @simonwillison.net · 21/05/2025

I got access to Gemini Diffusion, Google's first diffusion LLM, and the thing is absurdly fast - it ran at 857 tokens/second and built me a prototype chat interface in just a couple of seconds, video here: simonwillison.net/2025/May/21/...

simonwillison.net

Gemini Diffusion

Another of the announcements from Google I/O yesterday was Gemini Diffusion, Google's first LLM to use diffusion (similar to image models like Imagen and Stable Diffusion) in place of transformers. …

313518

Peter Bull @peter.drivendata.org · 21/05/2025

PSA to test suites with HTTP servers on Windows: localhost can be 100x slower than 127.0.0.1! 🤯 I hit the issue on a new cloudpathlib feature for HTTP that's coming soon github.com/drivendataor... Nearly impossible to debug. More background here medium.com/hackernoon/h...

medium.com

How changing ‘localhost’ to ‘127.0.0.1’ sped up my test suite by 1,800%

I would like to share a (somewhat recent) anecdote on how a one-line code change improved my understanding of software engineering.

000

Reposted by Peter Bull

Ethan Mollick @emollick.bsky.social · 20/05/2025

I say this a lot, but the narrative that AI use is going to collapse due to data limits or costs or environmental factors or regulation or a "hype bubble" popping or whatever is not a useful position for critics. Capability development may slow (it hasn't done so yet), but AI use isn't going away.

58312

Reposted by Peter Bull

Guillotine Hunger Force @handle.invalid · 17/05/2025

typical corrupt science

=
Nation & World -
The Seattle Times
My Account ™
The T. Rex may have been a lot smarter than you thought
Jan. 9, 2023 at 7:36 am | Updated Jan. 9, 2023 at 7:36 am
By DINO GRANDONI
The Washington Post

152155232308

Peter Bull @peter.drivendata.org · 13/05/2025

Chatbot benchmarks are proliferating and getting cited with every new release. We're thinking about the overlap between benchmarks and challenge leaderboards (www.drivendata.org/competitions/). Need to make better ones for social sector tasks to push model releases towards what matters.

drivendata.org

Competitions

DrivenData hosts data science competitions to build a better world, bringing cutting-edge predictive models to organizations tackling the world's toughest problems.

000

Peter Bull @peter.drivendata.org · 12/05/2025

AI alone won't change the social sector. I put down a few thoughts spurred at #GTF2025 on what other infrastructure—technical and social—we need to make AI work. Let me know your thoughts! www.linkedin.com/pulse/tech-i...

linkedin.com

(Tech) Infrastructure Week for the Nonprofit Sector

This Good Tech Fest didn’t start for me with “Data science is bullsh*t, right?” like last year, but the 2025 edition was just as provocative. After the conference, I’ve been working to crystalize…

011