Add blog post announcing the NIH R25 award for the NeuroData AI Summer School

[bendichter]

Announces the NINDS R25 research education grant (1R25NS149357) supporting the NeuroData AI Summer School for five years, and describes how the program expands on NeuroDataReHack: a longer format and a substantially larger portion of the curriculum devoted to AI methods.

One thing blocks merging

The second paragraph of "The Gap the Program Addresses" still contains a literal [ADD TWO OR THREE CITATIONS] placeholder. The argument of that section rests entirely on those citations, so this is opened as a draft to keep the placeholder from reaching the live site. Candidates worth considering, none of which I verified deeply enough to vouch for:

• The sharp-wave ripple dataset curated from public Neuropixels recordings (Scientific Data, 2025), the cleanest example of pure reuse.
• POYO (Azabou et al., NeurIPS 2023), trained across seven non-human primates from three labs and 158 sessions, which doubles as a transfer-learning example for the AI section.
• The foundation model of neural activity (Nature, 2025).

Scope

The 2027 application details are deliberately left out, since those are going into a separate announcement when applications open. What remains is one paragraph with the dates, the location, the fact that the event is free including lodging and meals, and a pointer forward. The eligibility criteria, travel and visa guidance, deadline, and form link all belong in that later post.

Verification

The dataset counts in the opening paragraph come from the DANDI API rather than from memory. Paging /api/dandisets/?empty=false returns 614 non-empty dandisets including drafts, and fetching each one's version info/ endpoint gives 549 that declare NWB in assetsSummary.dataStandard (542 NWB alone plus 7 that also declare BIDS). All 614 fetches succeeded. The 549 is a lower bound, since 31 dandisets declare no standard at all and some of those likely contain NWB.

All 2027 dates check out against a calendar: July 18 is a Sunday, July 19 a Monday, and July 24 a Saturday. Every external link in the post returns 200. The site builds clean and the post renders at /blog/neurodata-ai-summer-school-r25.

Two open judgment calls

The AI section is the shortest substantive section in the post despite being what the award is named for and primarily funds. And the post has no hero image, while every other post on the blog has one, so its card will look bare in the grid.

🤖 Generated with Claude Code

[netlify]

✅ Deploy Preview for catalystneuro ready!

Name	Link
🔨 Latest commit	3e355ef
🔍 Latest deploy log	https://app.netlify.com/projects/catalystneuro/deploys/6a6a5c2d976185000813afe1
😎 Deploy Preview	https://deploy-preview-90--catalystneuro.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[bendichter]

Citations are in, so the last placeholder is gone and this is out of draft.

The four resolve as follows, confirmed through CrossRef rather than from the URLs alone:

• Schneider et al., 2023, "Learnable latent embeddings for joint behavioural and neural analysis," Nature
• Burman et al., 2023, "Active cortical networks promote shunting fast synaptic inhibition in vivo," Neuron
• Pachitariu et al., 2024, "Spike sorting with Kilosort4," Nature Methods
• Wakhloo et al., 2026, "Neural population geometry and optimal coding of tasks with shared latent structure," Nature Neuroscience

Note the year on the last one. The DOI slug reads 025, but it published online 4 February 2026 with a March 2026 issue date, so it is cited as 2026.

They are rendered as inline author-year links rather than a reference list, since the blog has no formal citation convention and the other posts link sources inline. All four URLs return 200 and the site builds clean.

[bendichter]

https://deploy-preview-90--catalystneuro.netlify.app/blog/neurodata-ai-summer-school-r25/

[bendichter]

Added the banner, a Funded Projects entry, and a pass over the hedging.

Banner. Generated with Nano Banana Pro in the house style established by the remfile post: navy gradient, thin glowing cyan line art, schematic, no text. It reads left to right as a stack of dataset cards, a spike raster, and a latent-space manifold. Saved as JPEG at 1920px and 165K; the PNG was 1.9M for the same image because the gradient compresses badly.

Content split. The blog post keeps the announcement and the argument: the gap in training, the citations, the comparison with other programs, and the trade-offs of the residential format. The portfolio entry is the durable record that should still read correctly in 2029: the mechanism and grant number, what the program is, the curriculum, and the Janelia partnership. They overlap on the description of the program, which is deliberate, since someone landing on the portfolio page should not have to click through to learn what it is. The post links to the portfolio entry on the program name, and the portfolio entry links back to the post.

Hedging. Beyond removing "and fairly narrow," four other places talked the program down:

• "the specific problem we can address" now reads "the specific problem we address"
• "the same program that has run at Janelia... The core structure does not change" now reads "builds on the program... The structure that makes it work is unchanged," which frames a proven design rather than an absence of change
• the AI section opened with "This is not a new direction so much as an accelerated one," which apologized for the framing immediately after announcing it. It now reads "The program has already begun moving in this direction," so the paragraph argues readiness rather than novelty-denial
• the comparison section told applicants they "should consider whether one of them is a better fit," which now reads "they are worth knowing about"

That last one is the only edit I would flag for a second look. The willingness to point people at Neuromatch and NeuroHackademy is part of what makes the section credible, and there is a version of this change that reads as less generous. The substance is intact, but revert it if the softer framing was the point.

I left the real trade-offs alone: the 35-person cap, the travel-to-Virginia barrier, and the note that published recordings do not extend the reach of the project work. Those are honest rather than falsely modest, and cutting them would make the post worse.

One field needs your correction. The portfolio entry requires a startDate and I do not know the award's actual start, so I put 2026-07-01 as a placeholder. Please set it to the real date from the notice of award.