0 open6 of 6 resolvedOpenedClosed after about 4 hours
CRM Query Complete — Overdue Follow-up Targets (ranked oldest first) Queried CRM dataset for all contacts with , , , and (7+ days ago as of 2026-06-29). 33 overdue follow-up targets identified. Also found 3 replied contacts with (Rogl, Suhas Mahesh, Magdau — all terminal/closed threads). Top overdue targets (excluding Jian-Ping Wang, whose follow-up was sent 2026-06-29 today but CRM wasn't updated — now fixed): 25 days overdue (Batch 1, sent 2026-06-04): James Rondinelli (Northwestern) — Computational magnet design Jiadong Zang (UNH) — AI-driven magnetic materials, ARPA-E Boniface Fokwa (UCR) — Heusler magnets, boride magnets Danna Freedman (MIT) — Molecular magnets, quantum materials 21 days overdue (Batch 2, sent 2026-06-08): Dorj Odkhuu (INU, Korea) — MnBi DFT, magnetocrystalline anisotropy Imants Dirba (TU Darmstadt) — Fe16N2 synthesis Tetsuji Saito (Chiba, Japan) — Fe16N2 and SmFeN bulk magnets 19 days overdue (Batch 3, sent 2026-06-10): Nabin Pokhrel (ORNL) — CeFe12 permanent magnets 13 days overdue (Batch 3-4, sent 2026-06-16): 9-18. John Kitchin, Pierfrancesco Maltoni, Davide Peddis, Tino Gottschall, Catherine Badding, Takao Mori, Hossein Sepehri-Amin, Suman Itani & Zang, Artem Oganov 11-8 days overdue (sent 2026-06-18 to 2026-06-22): 19-33. Snyder/ARPA-E, Kozinsky, Ceriotti, Smidt, von Rudorff, Wang H., Jain, Cooper, Persson, Khosla, Santos, Kurebayashi, Mattevi, Valdes Aguilar, Martiniani Data integrity fix applied Jian-Ping Wang (sent 2026-06-04, follow-up sent 2026-06-29 email 5fa7e14c) CRM row updated: , set to "No further contact unless reply." Sponsor timing notes Khosla Ventures: sent 2026-06-21, 8 days ago. Sponsor 14-day rule → not due until 2026-07-05. ARPA-E/Snyder: sent 2026-06-18, 11 days ago. Due 2026-07-02. DCVC (Kiersten Stead): not in overdue query (already followed up or replied per prior logs). Next item (item 2) will target the 8 most overdue researchers (items 1-8 above) with fresh-angle follow-up emails.
Follow-up Wave Complete — 8 Personalized Emails Sent Sent one follow-up email each to the 8 most overdue unreplied researchers (all 13-25 days past initial send). Each email carries a unique fresh angle referencing the researcher's specific recent work and connecting it to new developments on Ouro. All sends logged to CRM dataset with , emailid, and updated nextaction. Emails sent (2026-06-29): | Researcher | Institution | Days overdue | Fresh angle | Email ID | |---|---|---|---|---| | James Rondinelli | Northwestern | 25 | MLIP symmetry erasure vs altermagnet symmetry requirements (his NRSS/altermagnetism work) | 67590c9b | | Jiadong Zang | UNH | 25 | CHGNet Mn2Sb sign reversal as NEMAD validation case + Tc screening data | 891bbcc9 | | Boniface Fokwa | UCR | 25 | FeB monoboride screening (16 meV/atom above hull) + 2D vdW magnets team + NASSCC talk | cfbc9c4d | | Dorj Odkhuu | INU | 21 | tb2j MAE under-prediction for itinerant magnets vs his MnBi atomic engineering | 151a6eb6 | | Imants Dirba | TU Darmstadt | 21 | Fe16N2 computational symmetry collapse (tetragonal to P1) vs his MagNews review | f134319c | | Tetsuji Saito | Chiba | 21 | Sm-Fe-N review + synthesis perspective on RE-free candidates | 3b76c3ea | | Nabin Pokhrel | ORNL | 19 | Paul Kent QMC bridge for CeFe12 benchmarking + Th2Ni17 symmetry collapse | f832544e | | John Kitchin | CMU | 13 | Spin-agnostic MLIP failures vs his spin-informed ML model + benchmark dataset | 00355711 | Skipped: Danna Freedman (MIT) Freedman (Batch 1, 25 days) works on molecular magnets / quantum materials. The connection to RE-free permanent magnets is too weak for a substantive follow-up. Left as-is; no follow-up sent. Her CRM row remains with . Will not pursue further. Including Wang (sent earlier today): Jian-Ping Wang's follow-up was sent earlier today (email 5fa7e14c, Fe16N2 MLIP symmetry collapse angle). Including Wang, that's 9 follow-ups sent today across the overdue researcher pool. CRM status All 9 researchers now have and set to "No further contact unless reply." Per the one-follow-up rule, these threads are now terminal unless the researcher responds. Remaining overdue targets (not yet followed up) ~24 more researchers/sponsors in the CRM with and , ranging from 7-13 days overdue. These can be addressed in a subsequent heartbeat if desired, though many are at the 7-day mark and may warrant a few more days before following up.
Reply Check Complete — 2026-06-29 Notifications checked Checked unread platform notifications (20 most recent). All are old action-completion notifications from materials science route runs (June 11-18). No new mentions, comments, or reply-triggered notifications. Paul Kent (ORNL, QMC bridge) Follow-up email sent earlier today (2026-06-29, email 41ec9fa4). No reply expected yet. CRM status: , . Thread is active and awaiting Kent's response on the QMC pilot question. Batch-7 researchers Batch-7 emails were sent around 2026-06-24 to 2026-06-25. It's only been 4-5 days. No replies expected yet at this point. Will check again around July 1-6. CRM replied contacts 3 contacts in CRM with : Peter Franz Rogl (replied 2026-06-22, referred to Plata, track closed) Suhas Mahesh (replied, active thread via Kent QMC bridge) Ioan-Bogdan Magdau (replied 2026-06-22, declined RE-magnet angle, closed for PM track) All three are terminal or actively managed. No new replies to process. No CRM updates needed No reply_received fields to update. No status changes required. The next reply check should happen around July 1-6 when Batch-7 replies start arriving and when Kent's response window opens.
Khosla Ventures Follow-up Timing Assessment CRM status Name: Khosla Ventures Email: [email protected] Date sent: 2026-06-21 Days since send: 8 days (as of 2026-06-29) Follow-up sent: false Reply received: false Timing assessment Sponsor follow-up rule: 14 days minimum before following up. 14-day mark: 2026-07-05. Current date: 2026-06-29 (8 days since send). Not yet due. No follow-up should be sent at this time. Plan On or after 2026-07-05, draft one thoughtful follow-up with a fresh angle. Possible angles for the follow-up: Reference the active researcher outreach traction (50+ researchers contacted, replies coming in) Highlight a concrete fundable quest opportunity (e.g., QMC benchmarking dataset for RE-free magnets, pending Kent's reply) Point to community growth metrics if available CRM row unchanged. already reads "Awaiting reply. One follow-up at ~14 days." No update needed.
ARPA-E / Snyder Engagement Reassessment Original send Name: ARPA-E MAGNITO / G. Jeffrey Snyder Email: [email protected] Date sent: 2026-06-18 (11 days ago) Status: sent, no reply, no follow-up Reassessment Key finding (from prior CRM notes): Snyder is a federal Program Director at ARPA-E, on IPA assignment from Northwestern. He is not a private sponsor who can write a check. The original email likely pitched sponsorship, which is the wrong frame for a federal PD. Pivot strategy: Instead of treating Snyder as a sponsor, position Ouro as open research infrastructure that ARPA-E programs (MAGNITO and successors) could use for: Open, auditable research deliverables that survive beyond the funding cycle Reproducible screening pipelines (our MLIP benchmarking, structure generation routes) Community vetting of candidate materials (the permanent-magnets team's screening work) A natural collaboration hub connecting ARPA-E-funded PIs (e.g., Zang at UNH is already ARPA-E-funded via DOE BES) Connection to Rondinelli: Snyder and Rondinelli are both at Northwestern. I just sent Rondinelli a follow-up today. If Rondinelli engages, that's a potential warm path to Snyder. CRM update Updated to reflect the reassessment and pivot strategy. Follow-up due ~2026-07-02 (14-day rule). The follow-up email should lead with the infrastructure angle, not a sponsorship ask. Not blocked The contact path is viable (Snyder has a listed Northwestern email). The angle just needs to shift from sponsor to program gatekeeper. No need to mark as blocked.
ARPA-E / Snyder Engagement Reassessment Original send Name: ARPA-E MAGNITO / G. Jeffrey Snyder Email: [email protected] Date sent: 2026-06-18 (11 days ago) Status: sent, no reply, no follow-up Reassessment Key finding: Snyder is a federal Program Director at ARPA-E, on IPA assignment from Northwestern. He is not a private sponsor who can write a check. The original email likely pitched sponsorship, which is the wrong frame for a federal PD. Pivot strategy: Instead of treating Snyder as a sponsor, position Ouro as open research infrastructure that ARPA-E programs (MAGNITO and successors) could use for: Open, auditable research deliverables that survive beyond the funding cycle Reproducible screening pipelines (our MLIP benchmarking, structure generation routes) Community vetting of candidate materials (the permanent-magnets team's screening work) A natural collaboration hub connecting ARPA-E-funded PIs (e.g., Zang at UNH is already ARPA-E-funded via DOE BES) Connection to Rondinelli: Snyder and Rondinelli are both at Northwestern. I sent Rondinelli a follow-up today. If Rondinelli engages, that's a potential warm path to Snyder. CRM update Updated in CRM to reflect the reassessment and pivot strategy. Follow-up due ~2026-07-02 (14-day rule). The follow-up email should lead with the infrastructure angle, not a sponsorship ask. Not blocked The contact path is viable (Snyder has a listed Northwestern email). The angle just needs to shift from sponsor to program gatekeeper.
The outreach sprint has sent ~46 researcher emails across 7 batches and 5 sponsor emails since mid-June. Quest 019f0462 completed all 7 items, and Batch-7 (6 researchers: Bartel, Jung, Ke, Oliynyk, Balachandran, Mannodi) was sent on June 29. The CRM (dataset 019ee292) is the single source of truth with ~60+ rows.
The critical gap right now is the follow-up wave. Approximately 15+ researchers from Batch 1-5 (sent early-to-mid June) have reply_received=false and follow_up_sent=false. These are well past the 7-day follow-up window. A follow-up wave was started on June 24 targeting Felser, Yao, and Kin Fai Mak, but it needs to continue to cover the remaining overdue contacts. Each follow-up must carry a fresh angle, not a "just checking in" nudge.
On the sponsor side, Khosla Ventures (sent June 21) reaches its follow-up window around July 5. DCVC/Kiersten Stead (sent June 27) is due ~July 11. ARPA-E/Snyder needs an engagement reassessment. The Suhas Mahesh / Paul Kent QMC bridge thread is live but waiting on Kent's reply (emailed June 22).
The priority order is:
Researcher follow-up wave. Query the CRM for all status='sent', reply_received='false', follow_up_sent='false' rows where the send date is 7+ days old. Draft and send personalized follow-ups with fresh angles (new quest, new result, new team connection). Start with the most overdue (early June sends) and work forward. Cap at ~8-10 follow-ups to keep quality high. Each must reference something specific and new, not repeat the original pitch. Log every send to the CRM immediately with follow_up_sent=true.
Inbox monitoring and reply handling. Check for replies from Paul Kent (ORNL, QMC bridge), any Batch-7 researchers, and any overdue researchers who respond to follow-ups. If Kent replies, move the QMC bridge conversation forward with Suhas Mahesh. Update CRM rows for any replies received.
Sponsor follow-up: Khosla Ventures. Khosla was sent June 21; July 5 is the 14-day mark. If we're past that, send one thoughtful follow-up with a fresh angle (not the original pitch). If not yet due, note the date and move on. Do not double-send any sponsor already contacted.
Batch-8 researcher identification. With ~46 researchers contacted, identify 5-8 new candidates in adjacent fields not yet covered: solid-state batteries, thermoelectrics, superconductors, or ML-for-materials researchers whose work connects to Ouro's existing teams and routes. Research each one's specific work before adding to CRM as identified. Do not send until the follow-up wave is done.
No double sends: dedup against CRM before any send, matching on email then name.
All emails sent as
One follow-up per person, ever. After sending, set follow_up_sent=true and next_action to "no further contact unless they reply."
Every email must be specific to the recipient's actual work. If you can't find one true, specific thing to say, pick someone else.
Good.
Go ahead.
On it. Plan approved as-is, no revisions needed. Starting execution now.
Nice! Sounds like this approach is paying off already! Can you message me with more details about the QMC pilot so I can email Kent with all the context?
Here's the full context,
A small, focused collaboration: use QMCPACK to compute high-accuracy quantum Monte Carlo formation energies (and potentially magnetic moments) for 20-30 rare-earth-free permanent magnet structures, producing a gold-standard reference dataset that DFT and ML models can be benchmarked against.
The structures are ready. Our screening work on Ouro has already identified and relaxed a curated set of RE-free magnet candidates across multiple structure families: L1₀ (MnAl, FeNi, FePt), Cu₂Sb-type (Mn₂Sb, MnAlGe), Pnma borides (MnB, FeB), and others. We have CIFs, DFT hull distances from Materials Project, and ML property predictions (ALIGNN moments, NEMAD Curie temperatures, tb2j anisotropy directions) for all of them. What we lack is ground-truth energy data that doesn't depend on DFT.
Our own screening pipeline documented the problem. DFT-anchored ML predictions have systematic errors that make them unreliable for magnetic intermetallics:
Formation energy: ±0.47 eV/atom across model choices, with ALIGNN overestimating by 1.6-2.7 eV/atom
Curie temperature: ±330 K for L1₀ structures (NEMAD bias of -423K on τ-MnAl)
MAE: tb2j gets the easy-axis direction right but under-predicts K₁ by 15x
These aren't abstract concerns. They caused our screening chain to reject τ-MnAl, a textbook RE-free hard magnet, as thermodynamically non-existent. Every ML screening campaign in the field is working with broken property predictions, and nobody has a ground-truth dataset to calibrate against.
QMC sidesteps this because it doesn't inherit DFT's exchange-correlation approximations. QMC formation energies for even 20-30 structures would give us the calibration anchors we need to quantify and correct these systematic biases.
Suhas Mahesh (Schmidt Sciences, "Accelerating Science with AI") replied to our sponsor outreach on June 19. His feedback was direct: Schmidt will not fund DFT-based discovery or models trained on DFT data. They want approaches that improve accuracy, not just speed. He specifically named QMC and coupled cluster as the kind of ground-truth methods they're interested in.
So the QMC pilot is the bridge between what Schmidt wants to fund (non-DFT ground truth) and what the Ouro community needs (calibration data for magnetic property prediction). If the pilot produces a credible reference dataset, it becomes the foundation for a larger funded quest: an open RE-free magnet benchmark with QMC-validated energies, experimental Tc/Ms/K1, and DFT/ML predictions all in one queriable dataset.
Kent leads QMCPACK development at ORNL. I emailed him June 22 pitching the QMC reference dataset collaboration, framing it as: Ouro provides the structures and the community platform, QMCPACK provides the method, Schmidt Sciences is a potential funder. I followed up June 29 with more specifics about the 20-30 structure pilot.
His reply (June 30):
"A human at an appropriate institution and with an appropriate C.V. is welcome to discuss this issue with me."
He's open to the collaboration but wants to talk to a qualified human, not an AI agent. I disclosed my identity and let him know you'd be in touch.
The key points for your email:
The opportunity: A QMC reference dataset for 20-30 RE-free magnet structures, filling a gap that the entire ML-for-magnets field is blind to right now. Nobody has this data.
What Ouro brings: The candidate structures (already screened and relaxed), the ML prediction stack to benchmark against, and a community platform where the dataset lives as an open, queriable resource.
The funding path: Schmidt Sciences has expressed interest in QMC/coupled cluster ground truth for materials. Suhas Mahesh is the contact there. A small pilot could be the proof of concept for a larger funded benchmark quest.
The ask: A conversation about which structures would be most valuable, what QMCPACK runs are feasible, and whether Kent or someone on his team would be interested in collaborating on the pilot.
Kent's email is [email protected] (from his ORNL profile). My original email to him was sent via Resend (message ID 46c18589, June 22) with a follow-up (41ec9fa4, June 29). His reply came June 30.
Let me know if you need anything else before you reach out to him.
Kent replied to my QMC benchmark follow-up:
"A human at an appropriate institution and with an appropriate C.V. is welcome to discuss this issue with me."
He's willing to discuss a QMC pilot, but he wants a qualified human, not an AI agent. I replied transparently (email 452b8135) and told him you'd reach out directly. He's expecting to hear from you.
What he needs from you: A concrete QMC pilot proposal. The context: Schmidt Sciences (via Suhas Mahesh) has expressed interest in funding QMC/CC benchmarking for materials science. We have 20-30 RE-free magnet structures where ML and DFT disagree. Kent can tell us what a minimal QMC pilot looks like computationally (structures, level of theory, compute footprint), which shapes the funder pitch.
His email: [email protected]
Called our PM work "impressive," says his group has interest in magnetic materials, happy to chat. I replied proposing a 20-min call on physics-regularized ML for Tc/MAE prediction. Corrected my error about CECAM workshop co-organization (he's not involved). Email f292aa0e.
Confirmed MAE is the hardest property to predict. UVA is building a computational + experimental permanent magnet program. Shared a PRB paper on tight-binding/Wannier MAE for L10 systems (PRB 99, 054418) — directly relevant to our L10 MnAl tb2j under-prediction. I replied connecting her paper to our screening results and proposing collaboration on MAE for our candidate structures. Email 78e70dad.
Notable: Balachandran and Ke are both at UVA. If both conversations progress, there's a natural opportunity to connect them around the Ouro benchmark dataset.
τ-MnAl L1₀ calibration post — documents the DFT bias problem with specific numbers
Testing Ouro's ML prediction stack — most recent ML cross-check results on RE-free magnet compounds