Our Projects
Encouraging Second-Look Case Reviews
In January 2025, the Office of the State Public Defender (OSPD), in partnership with Redo.io, sent letters to Public Defender’s Offices across California’s 58 counties containing the names and case IDs of potential candidates for second-look review. Working with the Three Strikes Project at Stanford Law School, we designed the selection criteria and the definition of “non-non-non” offenders: individuals serving time for offenses that are non-violent, non-serious, and non-sexual. We then generated a cohort of candidates for each sentencing county from our open database of more than 90,000 prison sentences and referred those county-specific lists to public defenders for review.
Depending on the case, relief may be available through several channels: recall and resentencing under Penal Code 1172.1, review of sentences ordered by the Board of Parole Hearings under AB 600, prosecutor-initiated resentencing, or a claim under the California Racial Justice Act where sentencing disparities support one.
California has built an expanding set of second-look mechanisms over the past decade. In 2018 it passed the first Prosecutor-Initiated Resentencing (PIR) law in the country, empowering district attorneys to revisit past cases, and supported it with an $18 million pilot across nine counties, independently evaluated by the RAND Corporation. Subsequent legislation, including AB 600, broadened the courts’ authority to recall sentences, and the Racial Justice Act opened a separate pathway for relief where a sentence was influenced by racial bias.
Yet a persistent “second-chance gap” separates legislative intent from implementation. Because these laws are discretionary rather than mandatory, outcomes vary dramatically across California’s 58 counties. Some jurisdictions actively pursue post-conviction relief; others exercise their discretion minimally or not at all. Only nine counties participated in the PIR pilot, and as of early 2025, roughly 200 individuals had been resentenced through it, far below expectations. The gap is not unique to any one statute: whichever channel applies, someone must first identify the cases worth reviewing.
The Challenge
Public defenders face substantial barriers in identifying eligible cases:
- Limited technology budgets (public defenders receive significantly less funding than district attorneys)
- Nearly half of California counties lack institutional public defender offices, relying instead on fixed-rate contracts that typically exclude post-conviction work
- Traditional case review requires expensive data scientists or contractors for statistical analysis
- Weeks or months to obtain data through California Public Records Act (CPRA) requests
Our Approach
Together with the Three Strikes Project, we developed eligibility criteria and a working definition of non-non-non offenders. Applying these criteria to our database covering the CDCR population, we identified individuals sentenced to more than 20 years who have served at least half their sentence, broadly excluding homicide, sex offenses, and super-strike convictions.
We then organized the results into cohorts by sentencing county, so that each Public Defender’s Office received a list specific to its own jurisdiction, and referred those lists through OSPD for attorney review where applicable.
Outcomes
Because the cohorts cover every sentencing county, candidates are identified across all jurisdictions, including the 49 counties that did not participate in the PIR pilot and those without institutional public defender offices. To date, working with OSPD, we have referred over 450 cases to public defender offices for review. Which legal vehicle fits a given case, if any, is a determination the attorneys make; our role is to make sure the candidates surface.
For attorneys who wish to conduct deeper analysis, our AI-powered platform offers advanced capabilities to:
- Instantly search and analyze prison records through natural language queries
- Identify similarly situated defendants using vector-based analytics
- Generate statistical evidence of racial bias for Racial Justice Act claims
- Assess rehabilitation progress and case suitability
- Produce court-ready evidence for prima facie and discovery motions
Limitations
Since Redo.io’s database is constructed using the California Public Records Act (CPRA), some information relevant to candidate identification is unavailable due to privacy restrictions in the law:
- Charging data and arrest records
- Juvenile offense identification
- Health records and rehabilitation milestones (COMPAS scores, mental health treatment, educational achievements)
- Some expunged or sealed offenses
As a result, individuals on our eligibility lists require thorough case file review to confirm suitability, as disqualifying factors not disclosed under CPRA may exist. Our recommendations are designed to accelerate case discovery, enabling attorneys to focus their time on comprehensive legal analysis of the most promising resentencing opportunities.
Code and Criteria
How we build the referral lists: selection logic and cohort generation, in the open.
The framework that builds county-level candidate cohorts
Selection criteria and scenario definitions used for candidate identification.
References
Wright, R. F., & Levine, K. L. (2025). Legislatures and localized resentencing. Journal of Criminal Law and Criminology, 114(4), 797. https://scholarlycommons.law.umich.edu/faculty-articles/306/
Chien, C. V. (2020). America’s paper prisons: The second chance gap. Michigan Law Review, 119(3), 519. https://repository.law.umich.edu/mlr/vol119/iss3/3
