UW Psychiatry Residency · Research Chief Session

Keeping up.

Staying on top of the literature, finding research opportunities, and never again discovering a deadline in the past tense.

Robert Chen, MD, PhD  and  Lynn Ren, MD, PhD
Research Chiefs, UW Psychiatry  |  20 August 2026

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it1 / 25

Before anything else

This deck was built with an AI. Here is exactly what that means.

What Claude did

Assembled these slides, drafted the lab summaries from each unit's public website, gathered and formatted every link, and wrote the calendar you'll see later.

What was checked

Every link opened by hand. Every PubMed number on the next slide run live against NCBI today: measured, not estimated. Conference dates put through five independent verification rounds.

What may still be wrong

Lab descriptions were not confirmed with the PIs. 79 of 223 conference dates are projections from past cycles, not announcements. An AI can be confidently, fluently wrong.

Everything here is a starting point, not a citation. Each item links to its primary source. Before a deadline matters to you, open the link.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it2 / 25

The first problem

You cannot read it all. Here is the actual arithmetic.

Measured against PubMed on 20 August 2026, counting the trailing 365 days.

4,829
new records enter PubMed every day
324
of those are psychiatry-related, every day
57
appear each week across the 12 leading psychiatry journals

Fifty-seven a week is a reading list. Three hundred a day is a nervous breakdown. The whole game is getting from the second number to the first.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it3 / 25

The second problem

And the papers are the easy part.

The field keeps moving

Guidelines get revised. Drugs get approved. The specialty argues with itself about what it even is, and that argument doesn't appear in any journal's table of contents.

The opportunities are invisible

Travel awards, abstract deadlines, fellowships, poster sessions. None of it arrives in your inbox. You find out in November that the abstract was due in September.

Nobody sends you that email. So today we build the things that do.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it4 / 25

Part one

Staying on top of the research

Four channels, ranked by how much they give you back for the time they cost.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it5 / 25

The map

Four channels. They do different jobs.

Push: PubMed alerts

Tells you what came out. Complete, free, unranked by anyone's taste but yours. This is the backbone.

Audio: podcasts

Tells you what it means clinically. Fits in the commute, which is the only reason it actually happens.

Opinion: Substack

Tells you what to think about it, and who disagrees. This is where the arguments about the field live.

Automated: scheduled LLM jobs

Tells you what you asked to be told, daily. Highest leverage, highest error rate. Use with a hand on the brake.

Start with one. Start with the first one.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it6 / 25

Channel one

PubMed alerts are the highest-yield four minutes in this deck.

Set it up once, in residency, and it keeps working through fellowship and faculty.

It is complete

No algorithm decides what you see. If the journal published it, the alert contains it.

It is free, and it is yours

No paywall, no engagement optimisation, no feed. It arrives in email and then it stops.

You choose the journals

Which means you are choosing your own reading list rather than inheriting someone else's.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it7 / 25

Follow along · step one of four

Sign in, so PubMed has somewhere to put the alert.

  1. Go to pubmed.ncbi.nlm.nih.gov and click Log in, top right.
  2. Sign in through your institution. Pick University of Washington and use your NetID. You can also use Google or an NCBI account.
  3. Once you're in, your username replaces the Log in button. Clicking it opens Dashboard, which is where saved searches live.
PubMed home page with the account menu open, showing Dashboard, Publications, Account settings and Log out

The account menu, once you're signed in. Dashboard is the one you want.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it8 / 25

Follow along · step two of four

Paste this into the search box.

"JAMA Psychiatry"[ta] OR
"The American Journal of Psychiatry"[ta] OR
"The Lancet. Psychiatry"[ta] OR
"World Psychiatry"[ta] OR
"Molecular Psychiatry"[ta] OR
"Biological Psychiatry"[ta] OR
"Neuropsychopharmacology"[ta] OR
"Schizophrenia Bulletin"[ta] OR
"Psychological Medicine"[ta] OR
"The Journal of Clinical Psychiatry"[ta] OR
"BMJ Mental Health"[ta] OR
"Nature Mental Health"[ta]

What [ta] means

"Title abbreviation", meaning the journal field. PubMed accepts the full title or the NLM abbreviation and resolves both to the same journal.

What this costs you

About 57 papers a week. Titles and abstracts only, so it is roughly a five-minute skim.

Too much?

Keep only the first four journals plus BMJ Mental Health. That lands at 12 a week and is a perfectly respectable place to start.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it9 / 25

Follow along · step three of four

Turn the search into a standing alert.

  1. Run the search, then click Create alert. It sits directly under the search box.
  2. Name it something you'll recognise in an inbox. "Psychiatry, core journals" beats "search 1". Up to 100 characters.
  3. Set the schedule to Weekly. Daily is how people end up filtering these straight to archive.
  4. Set format to Abstract and max items to 50, then Save. Abstracts let you triage without opening anything.
Schedule
daily · weekly · monthly
Format
Summary · Abstract · Summary text · Abstract text · PubMed
Items
1 · 5 · 10 · 20 · 50 · 100 · 200

Every one of these is changeable later, so don't agonise. The only irreversible mistake is not making one.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it10 / 25

Follow along · step four of four

This is what you're building toward.

The My NCBI Saved Searches panel listing four saved searches with What's New counts and Last Searched dates

My NCBI → Dashboard → Saved Searches.

The "What's New" column

A live count of what has appeared since you last looked. Click the number to open just those. Clicking it resets the counter.

Run as many as you like

One for your core journals, one for your own research topic, one for a method you're learning. They cost nothing to keep.

Four minutes, once. Then it runs for the rest of your career.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it11 / 25

Channel two

Podcasts: the ones that survive contact with a real schedule.

The Carlat Psychiatry Podcast

Chris Aiken, MD & Kellie Newsome, PMHNP

Fifteen minutes, clinically concrete, no industry funding. Post-tests earn up to 30 CME a year. If you pick one, pick this.

thecarlatreport.com

Psychiatry & Psychotherapy Podcast

David Puder, MD

Long-form and therapy-forward. Nearly 270 episodes, so treat the back catalogue as a reference shelf, not a queue.

psychiatrypodcast.com

NEI Podcast

Neuroscience Education Institute

Semimonthly, mechanism-heavy psychopharmacology with key opinion leaders. Good when you want the receptor-level story.

nei.global

Psychopharmacology Institute

Flavio Guzman, MD

Structured psychopharm teaching rather than news. A CME program more than a podcast; useful when you want depth on one drug class.

psychopharmacologyinstitute.com

PsychRounds

Drs. Tanner Hewitt, Larry Wang, Bradley Miller

Made by residents, for residents. Short, practical, pitched exactly at where you are rather than where an attending remembers being.

Apple Podcasts · Spotify

Psychiatry Boot Camp

Mark Mullen, MD

The fundamentals done properly: catatonia, capacity, agitation, risk. Built as intern prep; described in Academic Psychiatry.

psychiatrybootcamp.com

CME claims are the publishers' own. Check current accreditation before you count on the credit.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it12 / 25

Channel three

Substack: where the field argues with itself.

Psychiatry at the Margins

Awais Aftab, MD

The best single window into conceptual and critical psychiatry. Rigorous, generous to opponents, never cynical. Start here.

psychiatrymargins.com

Astral Codex Ten

Scott Alexander

A psychiatrist writing long-form on evidence, statistics and how to think. Not a psychiatry newsletter; often the sharpest thing about it.

astralcodexten.com

Rational Psychiatry

Thomas Reilly, MD

A monthly roundup of psychiatric research from an Oxford DPhil. If you want one email that tells you what mattered, this is it.

rationalpsychiatry.substack.com

Jonathan Shedler

Jonathan Shedler, PhD

Psychodynamic technique, written with a blade. Best read as a corrective when everything else you consume is a trial registry.

jonathanshedler.substack.com

Socratic Psychiatrist

Martin Greenwald, MD

Essayistic clinical reflection. Funnier than the others, and quietly serious about what the work does to the person doing it.

socraticpsychiatrist.substack.com

Make them a podcast

In the Substack app (not the website), tap the play icon at the top of a post and it reads itself aloud. Most publications have it. That turns this whole column into commute time.

English-language posts only

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it13 / 25

Channel three and a half

The best discussions of papers happen here. On both sides.

Nowhere else do authors, critics and methodologists argue about the same paper in public, within a day of it appearing.

None of that reaches you by default.

My view: worth it, as long as you know it is a standing cost and not a one-time setup.

Tailoring the feed is the whole job

Untailored, this channel is worthless. A feed worth reading takes days to weeks of following, muting and unfollowing, and there is no default that works.

And it never finishes

Stay vigilant. Drop your guard and the ragebait floods back in, because that is what the algorithm rewards once you stop correcting it.

Pay for the pro account

It costs money, and it buys no ads and a cleaner feed, which is what makes the tailoring actually stick.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it14 / 25

Channel four

Have an assistant read the internet at 6:30 every morning.

A scheduled-automations dashboard listing five active recurring jobs including daily psychiatry news and weekly drug-discovery summaries

My own scheduled jobs. Five running; 44 runs in the last 30 days.

What this is

A standing instruction to an AI assistant, on a timer. It searches, summarises and emails you. ChatGPT, Claude and Grok all do a version of this.

Why it earns its place

It covers what PubMed structurally cannot: preprints, FDA actions, conference news, trial registrations, and the trade press.

The brake

It will hallucinate a citation eventually. Treat every output as a pointer to check, never as a source to quote.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it15 / 25

Sixty seconds

Build one now.

A New Automation dialog with a name field, a weekly Thursday 6:30 AM trigger, an instructions box, and an Email plus App notification setting
  1. Name it so you know what it is in six months.
  2. Weekly, on a day you're not post-call.
  3. Write the instruction like you'd brief a medical student. Say what you want, what you don't, and how long.
  4. Send it to email, so it lands where you already triage.
Every Thursday, summarise the past week in
psychiatry: new RCTs, guideline changes, and FDA
actions. Ten bullets, each with a working link to
the primary source. Say "nothing this week" rather
than padding. Flag anything you could not verify.
Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it16 / 25

If you remember nothing else

Do these three. Total cost: about seven minutes.

  1. Set the PubMed alert

    Weekly, abstract format, the twelve journals we just pasted. Four minutes, once. It is the only channel here that is complete.

  2. Put Carlat in your podcast app

    Fifteen-minute episodes, clinically usable, free. Two minutes. It survives call weeks, which is the whole test.

  3. Subscribe to Psychiatry at the Margins

    Awais Aftab, roughly weekly, free tier is plenty. One minute. It is how you stay a thinking psychiatrist and not just an informed one.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it17 / 25

Over to you

What am I missing?

Every one of you has a source I don't. Newsletters, journal clubs, a fellow who forwards the good stuff, a listserv nobody documented.

Say it now and it goes on the resident page tonight.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it18 / 25

Intermission. Please do this now

Research Track Party at John Neumaier's house.

When
Saturday 19 September, 1:00 PM
Where
Shoreline, WA. Full address on the RSVP form
Host
Dr. John Neumaier
What
Food, drinks, science, and a real view of the Sound
Bring
Yourself, plus anyone you're bringing. The form asks

The form takes about thirty seconds: are you coming, your year, who's with you, and any dietary restrictions.

QR code linking to the Research Track Party RSVP form

RSVP now Point your camera at this.
forms.gle/QDTBfVnR295zDA7F6

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it19 / 25

Part two

Conferences, deadlines, and where to present

The part nobody hands you, and the part that costs a whole year when you miss it.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it20 / 25

The thing I built for this

A residency research calendar. Public, free, no account.

96
opportunities: conferences, travel awards, abstract deadlines, fellowships
223
individual dates on the timeline, out to the end of residency
138
confirmed against the organisation's own page, each with a link

Filters by your PGY year, subspecialty and research interests. Exports to your calendar. Runs entirely in your browser, so nothing you type leaves your machine.

QR code linking to the residency research calendar

Open it now We're going through it together.
therealryc.github.io/
residency-research-calendar

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it21 / 25

Follow along · live demo

Five minutes to make it yours.

  1. Open the link. A short wizard asks your PGY year, subspecialty interests and research areas. You can skip it, and skipping shows you everything.
  2. Watch the timeline filter. Ineligible opportunities drop away as you answer. Nothing is hidden permanently; you can always show everything again.
  3. Click any bar to open the detail panel: eligibility, cost band, notes, and the link to the organisation's own page.
  4. Export to your calendar. Choose which categories to include, then download an .ics and open it in Outlook or Google Calendar.

Where your profile lives

In your browser, on your machine. No account, no server, no email collected. Export a file if you want it on another computer.

Something missing?

Add your own deadline in the app, or open an issue on the repo. Winter Brain got added because someone asked.

github.com/therealRYC/residency-research-calendar

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it22 / 25

The important slide

How much to trust any given date.

Confirmed
138 dates
Read off the organisation's own page, with the link and the date it was checked. Solid.
Projected
79 dates
Next cycle isn't announced. This is last year's date rolled forward. Useful for planning, never for relying on.
Unconfirmed
6 dates
Something is there but we could not verify it. Shown so you know to look, flagged so you don't trust it.

Every single entry links to its source

That is the point of the whole thing. You should never have to take my word, or an AI's word, for a date that costs you a year.

Five verification rounds found a fabricated award, a deadline off by 71 days, and four stale cycles. Assume there is a sixth.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it23 / 25

Finding a mentor

44 psychiatry-related labs and centers at UW, in one page.

Compiled by Lynn Ren, MD, PhD, your other Research Chief. Search by disorder, method or name; filter by research type; click straight through to the lab.

44

labs, centers and programs

25

home units across UW

10

research-type filters

Descriptions were AI-drafted from each unit's public site and are not confirmed with the investigators. Directors change. Open the lab page before you email anyone.

QR code linking to the UW psychiatry labs and centers browser

Browse the labs Public, no login, works offline once open.
therealryc.github.io/
uw-psychiatry-labs

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it24 / 25

Everything from today

Three links. They all live on the resident page.

The calendar

96 opportunities, 223 dates, every one linked to its source.

therealryc.github.io/
residency-research-calendar

The labs browser

44 UW labs and centers, searchable. Compiled by Lynn Ren, MD, PhD.

therealryc.github.io/
uw-psychiatry-labs

The RSVP

Research Track Party, Saturday 19 September, 1:00 PM.

forms.gle/QDTBfVnR295zDA7F6

Set the alert before you leave the room. Everything else is optional.

Questions, additions, corrections: robchen@uw.edu. Corrections especially.

Built with Claude (Anthropic AI) · every claim links to its source. Verify before you rely on it25 / 25