Keeva Uses Client-Side Encryption and Walrus Memory to Persist a Credential Vault

·3 views

A developer has written up Keeva, a credential vault built for the Total Recall hackathon, which stores passwords, API keys, and private notes. The design encrypts records in the browser so that storage only ever receives ciphertext, with decryption possible only after explicit user authorization. Keeva divides responsibilities among a Sui wallet for authorization, Walrus Memory for persistent encrypted storage, and Tatum for blockchain reads and transaction checks. Encryption relies on AES-256-GCM with per-operation random IVs, PBKDF2-SHA256 key derivation, and context binding so that mismatched records fail to decrypt. The article also addresses harder persistence problems, such as password changes, conflicting edits across devices, and unacknowledged storage writes.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Un agente de IA logra 74 merges en 24 repositorios open source en seis semanas

Entre finales de septiembre y el 10 de noviembre de 2026, una cuenta operada por un agente de IA abrió 143 pull requests. De ellos, 74 fueron integrados, 49 de esos en 24 repositorios externos como browser-use y llama_index. Según el relato, cada PR se limitaba a una sola idea, llegaba con implementación, tests y CI en verde, y se respondía a los mantenedores en horas. Los autores también señalan que aceptaron los rechazos sin insistir, y que los PRs pequeños con tests fueron los que más rápido se aprobaron.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Daily Cognito Audit Job Could Silently Inflate AWS MAU Billing

A developer designing a multi-tenant SaaS with a single Cognito user pool planned a daily job to check for unexpected email changes and administrators lacking TOTP. AWS documentation states that AdminGetUser calls count toward monthly active user billing, so the job would make every account billable regardless of actual sign-ins. The author estimates costs on the Essentials tier, where each MAU beyond the 10,000 free allowance costs $0.015, and notes the cost scales with signed customers rather than usage. A shared free tier across an organization, including staging accounts, could push spending past the threshold earlier than expected. The author's planned safeguard, an AWS Budgets alert at 9,000 MAU, is unverified and may not function as intended.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Afterword: Measuring Which Blog Comments Actually Matter Using a Ledger of Claims

The author analyzed nine years of DEV data and found that the platform's dashboard metrics told them little about what they valued. They concluded that the most meaningful signal lay in reader comments, which often include corrections, extended arguments, and counterexamples that shape later writing. To test whether an AI model could help triage the comment stream, they recorded their hypotheses and falsification criteria in a documented ledger before writing any code. The project, called Afterword, is a local tool that organizes comments by priority without hiding or deleting any of them. Its success will be judged mainly by whether comments the author rates as consequential stay visible, rather than by raw accuracy, and performance will be compared against both full chronological reading and a simple keyword heuristic.

0 comments Read more at DEV Community

ProgrammingHacker News ·

U.S. Policies Push Young Scientists to Leave, Report Finds

A New York Times interactive feature examines how federal funding cuts are affecting early-career researchers in the United States. The piece reports that uncertainty over research grants is discouraging young scientists, including PhD candidates, from pursuing careers in the country. It describes some researchers considering or pursuing opportunities abroad as a result. The article was shared on Hacker News, where it drew moderate attention and discussion. The source text provided contains only the link and engagement metrics, so the specific figures and examples are drawn from the article's framing rather than its full text.

0 comments Read more at Hacker News