Library, topics, positions, published work.
Four steps, in the order researchers have always worked. The difference is what comes out the other end.
Titrate is pre-launch. What is described here is what it is being built to do, with Jeff Casebolt, PhD.
The central question
Does the barbell hip thrust produce superior gluteus maximus development — measured either as acute EMG activation or as chronic hypertrophy — compared to the back squat?
- Contreras et al.2015n=13supports
- Delgado et al.2019n=8mixed
- Plotkin et al.2023n=34contradicts
- Krause Neto et al.2025n=12mixed
- Vigotsky et al.2017n=—contradicts
- Kassiano et al.2024n=33supports
What you get
What is in a Titrate workspace.
When the first cohort opens and you sign in, this is what a workspace holds. Eight surfaces, in the order the work moves through them — from the papers that land in your library, however they got there, to the thing that goes out with your name on it.
01 · Your library
The papers you have read, as uploaded PDFs or citation-only records. Two ways in — upload the library you already keep, or start from a topic and accept or reject what Topic Finder proposes. Titrate does not ship a corpus and never decides what belongs in yours.
02 · Topic stacks
A stated question with the papers assembled against it. Stacks can sit inside stacks, because real questions nest.
03 · Cited opinions
Your position on a stack: research question, thesis, body, take-home — and a citation per claim that points back to the paper in your library.
04 · Articles and pieces
What you publish, drafted from an opinion so the evidence chain survives into the published version. One workspace can route pieces to more than one brand you write under.
05 · Courses
Modules built from your own opinions, so you can see which module rests on which paper when the literature moves.
06 · Digests
An issue you assemble and send, rather than one you receive from someone else's editorial team.
07 · Subscribers
The list is yours. It is a first-class record in your workspace, not an export from a platform that owns the relationship.
08 · Client questions
A question and your cited answer in one thread — the client can follow the citation back to the paper and check you.
And what is deliberately not in it
No chat window. No assistant that drafts your position. No confidence score, no consensus meter, no answer generated on your behalf. Every one of the eight surfaces above is a place to put your own judgment and keep the trail attached to it — which is the only reason any of it is worth publishing under your name.
Titrate is pre-launch, and these eight surfaces are at different stages of being built. Each one is backed by real schema in the repository today, which is why they can be described this precisely — but none of it is being offered as a finished product, and there is no trial. There is a waitlist.
The four steps
01
The library — the papers you have read, held as uploaded PDFs or citation-only records. Two ways in: upload what you already keep, or start from a topic and have Titrate propose candidates for you to accept or reject. Either way the paper becomes a record carrying its title, authors, journal, year, DOI, abstract and provenance, and a field that could not be retrieved says unknown rather than guessing.
02
The stacks — a stack is a question, and the papers you have read that bear on it. It holds design, sample size and how each paper cuts, so a disagreement stays visible instead of averaging out.
03
The positions — you write what you concluded and why: research question, thesis, argument, take-home. Each claim carries a citation that points at the paper in your own library alongside the formatted reference. The citation list is a column on the position itself, not a file you keep in step in another app — so the trail travels with the position, and the link it holds is a real reference to the paper rather than a string you retyped.
04
The published work — a position becomes an article, a course module, a digest issue, or an answer to a client. Under your name, with the evidence chain intact through publication.
Step 1 — how the papers get in
Two reasonable ways to ask the same question. One returns your papers; the other returns none of them.
Titrate is being built with two ways in. Upload the library you already keep — the folders of PDFs, itemised by topic, that some people have been building for twenty years. Or start from a topic and have Titrate propose the reading. This section is about the second one, and it opens with the unflattering part, because the unflattering part is the whole design problem.
Before anything else: how much of it can you actually open?
Stated first, because it decides what a result can even be. One contested question — does passing objective return-to-sport criteria reduce second-ACL-injury risk — asked of PubMed on the day this page was built:
Papers indexed on the question
269100%
"anterior cruciate ligament"[MeSH] AND "return to sport"[tiab]
Free to read somewhere
9736%
+ free full text[sb]
In the redistributable open-access subset
3613%
+ pubmed pmc open access[filter]
Sixty-four percent of that literature can only ever reach you as a citation. That is not a Titrate limitation waiting to be engineered away — no product in this category redistributes paywalled full text, and none of them tells you the proportion before you search. It is why a paper in Titrate is a PDF where a free copy exists and a citation where one does not, and why a citation-only record with fields marked unknown is a complete entry rather than a broken one. Three of Jeff Casebolt's own five publications sit in that 64%. He cannot hand you those either.
The worked example from the spec, executed
Topic Finder is specified around keyword refinement: a topic, a second keyword to narrow, a third to narrow further, plus a year range, an author and a journal. Here is the example written into the spec, run as four cumulative queries. The left column quotes the spec.
Refinement
Query, as it accumulates
Results
results (log10)
01
Anterior cruciate ligament (ACL)
Query, as it accumulates:"anterior cruciate ligament"[MeSH Terms]Results: 13,216
02
Female athletes
Query, as it accumulates:+ AND "female athletes"[Title/Abstract]Results: 114
03
Volleyball
Query, as it accumulates:+ AND "volleyball"[Title/Abstract]Results: 4
04
Surgical procedure — hamstring or patellar tendon
Query, as it accumulates:+ AND ("hamstring"[tiab] OR "patellar tendon"[tiab])Results: 1
Four sensible refinements and the set collapses to one paper. Worse, it collapses to the wrong one. The survivor is a 2013 musculoskeletal-modelling study of stop-jumps in recreational female athletes (PMID 23015067) — not a comparison of hamstring against patellar-tendon grafts, which is what the fourth keyword was reaching for, and outside the ten-year window the spec asks for. A narrowing chain does not tell you when it has narrowed past the answer. Tested live on the same day: add one more term until the set is empty and PubMed returns "No results were found" with no indication of which term killed it and no way to drop just that one.
And the same first keyword, expressed four ways
This next part is worth knowing whether or not you ever use Titrate. Ask for one concept in the ways a person reasonably would, and you get different literatures.
| How the term is expressed | Typed plainly — PubMed expands it across all fields | Selected as a subject heading — the anatomical structure | Selected as a subject heading — the injury | Selected as a subject heading — the reconstruction |
|---|---|---|---|---|
| Query actually run | anterior cruciate ligament | "anterior cruciate ligament"[MeSH Terms] | "anterior cruciate ligament injuries"[MeSH Terms] | "anterior cruciate ligament reconstruction"[MeSH Terms] |
| Results | 35,129 | 13,216 | 16,040 | 9,561 |
| Grindem et al.2016 | Grindem et al. is in the results for Typed plainly — PubMed expands it across all fields | Grindem et al. is not in the results for Selected as a subject heading — the anatomical structure | Grindem et al. is not in the results for Selected as a subject heading — the injury | Grindem et al. is in the results for Selected as a subject heading — the reconstruction |
| Kyritsis et al.2016 | Kyritsis et al. is in the results for Typed plainly — PubMed expands it across all fields | Kyritsis et al. is not in the results for Selected as a subject heading — the anatomical structure | Kyritsis et al. is in the results for Selected as a subject heading — the injury | Kyritsis et al. is in the results for Selected as a subject heading — the reconstruction |
| Losciale et al.2019 | Losciale et al. is in the results for Typed plainly — PubMed expands it across all fields | Losciale et al. is not in the results for Selected as a subject heading — the anatomical structure | Losciale et al. is in the results for Selected as a subject heading — the injury | Losciale et al. is in the results for Selected as a subject heading — the reconstruction |
| Webster & Hewett2019 | Webster & Hewett is in the results for Typed plainly — PubMed expands it across all fields | Webster & Hewett is not in the results for Selected as a subject heading — the anatomical structure | Webster & Hewett is in the results for Selected as a subject heading — the injury | Webster & Hewett is in the results for Selected as a subject heading — the reconstruction |
| Paterno et al.2022 | Paterno et al. is in the results for Typed plainly — PubMed expands it across all fields | Paterno et al. is not in the results for Selected as a subject heading — the anatomical structure | Paterno et al. is in the results for Selected as a subject heading — the injury | Paterno et al. is in the results for Selected as a subject heading — the reconstruction |
| Zhou et al.2024 | Zhou et al. is in the results for Typed plainly — PubMed expands it across all fields | Zhou et al. is not in the results for Selected as a subject heading — the anatomical structure | Zhou et al. is in the results for Selected as a subject heading — the injury | Zhou et al. is in the results for Selected as a subject heading — the reconstruction |
| Of 6 papers an expert kept | 6 of 6 | 0 of 6 | 5 of 6 | 6 of 6 |
The narrower query is the one a field picker invites you to build, and it loses every paper the expert kept — because the right heading was never the anatomical structure, it was Anterior Cruciate Ligament Reconstruction. Nothing on the face of either query tells you which one you just ran. PubMed does expose it: the Details chevron in your search history shows how your terms were translated, one click deep in a table most people never open. So this is a borrowed convention, not an invention. If a builder is going to put a topic picker in front of you, the least it owes you is the translation — in front of you, before you trust the set.
What was kept — an outcome, not a control
Topic Finder proposes. You dispose. These six were kept by someone who read them, and each carries an identifier you can resolve while you read this. They are also the exact input to a topic stack — which is where the rest of this page picks up. Finding and organising are one chain here, not two products.
| Paper | Year | Design | N | Bears on the question |
|---|---|---|---|---|
| Grindem et al.doi:10.1136/bjsports-2016-096031 · PMID 27162233 | 2016 | Prospective cohort | 106 | supports |
| Kyritsis et al.doi:10.1136/bjsports-2015-095908 · PMID 27215935 | 2016 | Prospective cohort | 158 | supports |
| Losciale et al.doi:10.2519/jospt.2019.8190 · PMID 30501385 | 2019 | Systematic review / meta-analysis | 4 studies pooled | contradicts |
| Webster & Hewettdoi:10.1007/s40279-019-01093-x · PMID 30905035 | 2019 | Systematic review / meta-analysis | — | mixed |
| Paterno et al.doi:10.4085/1062-6050-0608.21 · PMID 36638338 | 2022 | Prospective case-cohort | 159 | contradicts |
| Zhou et al.doi:10.7717/peerj.17279 · PMID 38699196 | 2024 | Systematic review / meta-analysis | 9 studies / 1,410 | mixed |
2 supporting · 2 contradicting · 2 mixed
The query is real, the counts are real and dated, the papers are real — the screen is not, and neither is the retrieval. Every number on this page belongs to PubMed, for the query printed beside it, re-run 3 August 2026 through the NCBI E-utilities API. Re-run any of them. Titrate has no index of its own and is not being built to search better than PubMed, which already does most of what this describes and is free. The destination is built — a paper record already carries its source type, authors, journal, year and DOI — and the retrieval is not: there is no search endpoint in the repository today. That is the state of it, written down so you can check it.
Step 1 — what a paper becomes
A field it could not find says so.
This is a real record, and it is a useful one because it is imperfect. The 2005 paper below genuinely never had a DOI assigned. Here is what that looks like on the screen.
Title
Preferred and Energetically Optimal Transition Speeds During Backward Human Locomotion
Authors
Hreljac, Imamura, Escamilla, Casebolt, Sison
Journal
Journal of Sports Science & Medicine, 4(4), 446–454
Year
2005
DOI
PMID
24501559
Open access
Yes — full text in PMC
Source
citation-only record
Provenance
Retrieved 2 Aug 2026 · PubMed, PMC, Unpaywall
The rule this record demonstrates
A missing field marked unknown is fine. A fabricated field is not. That is the rule the seed corpus was built under and it is the rule the ingest pipeline is being built to follow: extract with a real extractor first, fall back to resolving a DOI or PMID, and if neither returns a record, save it with the fields explicitly unknown. Never a plausible guess. The paper above is one of five peer-reviewed publications belonging to the person building Titrate, so the rule is being applied to his own record first.
Titrate is pre-launch. The paper-detail screen is designed, not built. The record itself is real — check the PMID.
Step 2, shown
A stack is a question, not a folder.
Step two above is the one that is hardest to picture, so here it is with real material in it — six papers on one contested question, carrying what a practitioner actually judges them by.
The central question
Does the barbell hip thrust produce superior gluteus maximus development — measured either as acute EMG activation or as chronic hypertrophy — compared to the back squat?
| Paper | Year | Design | N | Bears on the question |
|---|---|---|---|---|
| Contreras et al. | 2015 | Cross-sectional EMG | 13 | supports |
| Delgado et al. | 2019 | Cross-sectional EMG | 8 | mixed |
| Vigotsky et al. | 2017 | Methodological commentary | — | contradicts |
| Plotkin et al. | 2023 | RCT — MRI-measured hypertrophy | 34 | contradicts |
| Kassiano et al. | 2024 | RCT — ultrasound thickness | 33 | supports |
| Krause Neto et al. | 2025 | Systematic review / meta-analysis | 12 studies | mixed |
2 supporting · 2 contradicting · 2 mixed
Why the stack is the step that matters
Nothing above this row is unusual — every reference manager stores papers, and every practitioner has a topic they care about. The step that does not exist anywhere else is this one: a question stated at the top, and the papers underneath carrying enough about themselves that a disagreement survives contact with them. A folder cannot hold a disagreement. A question can.
Which is why step three works. You are not writing a position against a pile of PDFs — you are writing it against a stated question with the counter-evidence already in front of you.
Every paper listed here is real and independently re-verified on 2 August 2026 against PubMed, PMC, the DOI resolver, Unpaywall and an open citation-graph API. The screen is not. Titrate is pre-launch, and the topic-stack view is designed, not built — this is the specification, shown with real material rather than mocked up. Labelling how a paper bears on the question is a design commitment, not a shipped feature, and it is not what makes Titrate different.
Step 3, shown
The empty field is the product.
When you write a position against a stack, this is the record it becomes. Four fields of your judgment, and a citation rail that cannot be empty. Titrate fills in none of the four.
Research question
Does the barbell hip thrust produce superior gluteus maximus development compared to the back squat?
Carried from the stack, so a position always answers a stated question.
Thesis
Your position. In your words.
Nothing drafts this. There is no chat window, no suggested wording and no "generate" button — this field is the reason the product exists.
Body
The argument, claim by claim.
Each claim can carry the paper it came from.
Take-home
What you would actually do, and why.
This is the field a client, a referrer or a student reads first.
Status
draft
An opinion stays a draft until you say otherwise.
The citation rail — one entry per claim
Contreras, B., Vigotsky, A. D., Schoenfeld, B. J., Beardsley, C., & Cronin, J. (2015). A comparison of gluteus maximus, biceps femoris, and vastus lateralis electromyographic activity in the back squat and barbell hip thrust exercises. Journal of Applied Biomechanics.
doi:10.1123/jab.2014-0301 · PMID 26214739
The barbell hip thrust activates the gluteus maximus and biceps femoris to a greater degree than the back squat when using estimated 10RM loads.
Vigotsky, A. D., Beardsley, C., Contreras, B., Steele, J., Ogborn, D., & Phillips, S. M. (2017). Greater electromyographic responses do not imply greater motor unit recruitment and "hypertrophic potential" cannot be inferred. Journal of Strength and Conditioning Research.
doi:10.1519/JSC.0000000000001249 · PMID 26670996
Elevated electromyographic measurements during exercise do not necessarily indicate greater recruitment of muscle fibers or predict greater potential for muscle growth.
Plotkin, D. L., Rodas, M. A., Vigotsky, A. D., et al. (2023). Hip thrust and back squat training elicit similar gluteus muscle hypertrophy and transfer similarly to the deadlift. Frontiers in Physiology.
doi:10.3389/fphys.2023.1279170 · PMID 37461495
Nine weeks of hip-thrust-only or back-squat-only training produced similar gluteal cross-sectional-area growth, modestly but not significantly favouring the hip thrust.
Why this cannot rot
Each entry holds a foreign key to the paper in your own library alongside the rendered reference and your note on why that paper is attached to that claim. The citation list is a column on the opinion — citations, NOT NULL — so it is part of the record rather than a file you keep in step somewhere else. What the column does not do is decide for you that three citations are enough: an opinion with an empty list is legal, and keeping it honest stays your job. What it does is make the connection between a position and a paper into data rather than prose — which is the difference between a trail you maintain by discipline and one that is part of the thing you wrote.
The three citations above are real papers and the notes are the authors' own framing, quoted. The screen is not: Titrate is pre-launch and the opinion composer is designed, not built. This is the specification shown with real material, not a screenshot.
A workbench, not a chatbot.
Nothing here writes a confident paragraph for you to trust. There is no chat window. You keep the judgment; the tool keeps the trail back to the papers. It cannot hallucinate a citation because it does not generate citations at all — a claim points at a paper already sitting in your own library, or it points at nothing.
Questions
Does Titrate find the papers for me?
It is being built to do both. You can upload the library you already keep, and you can start from a topic and have Titrate propose candidates from the public indexes — a keyword, a second and third to narrow, a year range, an author, a journal, a defined number back, returned as a PDF where a free copy exists and a citation where one does not. What it will never do is decide for you. It proposes; you dispose. That is not a limitation being dressed up as a principle: a library you did not assemble is not one you can defend, and everything downstream on this page depends on you having chosen. Be clear about the state of it, though — the destination is built and the retrieval is not. The section above shows exactly what that means, with numbers you can re-run.
Is this a reference manager?
It sits after one. A reference manager stores papers and metadata. Titrate holds the position you took and the reasoning behind it, with the papers attached.
What does it actually look like today?
The three surfaces above — the paper record, the stack and the cited opinion — are specified, not shipped. Rather than mock up a screen and call it real, you are looking at the material they hold: real papers, real DOIs, checkable while you read. When the first cohort opens you can judge the screens themselves.
Do I have to already have a library?
No. Topic Finder exists so that a cold start is not a wall — most people do not arrive with twenty years of PDFs in folders. What stays true is that nothing lands in your library without you putting it there. Titrate does not ship with a corpus, does not curate a reading list and does not tell you which papers matter. It can go and get candidates, and then it stops and waits for you.
If it returns PDFs, where do they come from?
From places allowed to give them to you: the open-access copy on PubMed Central, a publisher's own free version, an author's accepted manuscript in a repository. Never Sci-Hub, never a shadow library, never a scraped copy of a paywalled article. Where none of those exists you get the citation and the DOI and you go through your own institutional access — the same route every legitimate tool in this category takes, for the same reason. If your first thought on reading "returns PDFs" was that this had to involve something dubious, that was the right instinct, and this is the answer to it.
Titrate is pre-launch.
Join the practitioner waitlistOne email when the first practitioner cohort opens. Nothing else.
