
Jonah Reed
Language runtimes and performance
Jonah covers programming languages, runtimes, and performance. His articles explain runtime behavior and the assumptions that make benchmarks useful or misleading.
Engineering guides
- What should we do when an inactive replication slot is filling the database disk?
When an inactive PostgreSQL replication slot is retaining WAL and eating disk, first decide whether the downstream consumer can resume, must be reseeded, or should be retired.
- How should permissions change when a record moves from draft to finalized?
A record moving from draft to finalized should not trigger a wholesale rewrite of permissions.
- How do we design deterministic floating-point behavior across CPUs, compilers, and runtimes?
Define floating-point reproducibility as a software contract, then constrain compilation, execution order, and numeric boundaries to match that contract.
- When should an image pipeline rotate pixels versus carry orientation metadata?
Normalize orientation at first decode when images will be cropped, resized, analyzed, or exported across mixed tools; keep metadata only when original bytes must stay untouched.
- How should an ADR record a decision with an explicit expiry date and revalidation trigger?
Record an ADR's expiry, measurable review triggers, owner, evidence, and exit path so temporary architecture decisions get revisited.
- Prevent OAuth Mix-Up in Multi-Provider Login
Bind each OAuth authorization response to the issuer chosen at login start, and require `iss` when supported to stop shared-callback provider confusion.
- Prevent Session Fixation at Login and Privilege Transitions
Rotate session identifiers at login, MFA, recovery, impersonation, and privilege changes, carrying forward only explicitly allowlisted anonymous state.
- Debug denied requests without logging sensitive policy data
Debug denied authorization decisions safely by logging stable decision facts instead of sensitive policy inputs, while separating developer diagnostics, audit records, and customer-visible events.
- Alternatives to replacing the identity provider in one global cutover
Can a product migrate by application, customer cohort, or credential event instead of switching every login at once? Compare bounded architecture and migration choices.
- OPA alternatives for document sharing and inherited access
Compare OPA with OpenFGA and Ory Keto on Ory Network for document sharing, nested groups, and inherited access.
- When DPoP Is Worth the Client and Verifier Complexity
Evaluate sender-constrained OAuth tokens by tracing keys, proofs, nonces, replay windows, proxies, and realistic token theft.
- Incident Command for Teams Without a Dedicated Commander
Divide mitigation, investigation, communication, and decisions across a small on-call team without creating collisions.
- What crash-safety contract should an application assume when replacing or appending a file after power loss?
A practical contract for crash-safe file updates: what `rename()` makes atomic for visibility, what `fsync()` makes durable, and what recovery must still verify.
- When to use a hard cutover vs. a transition window for enterprise SSO
How to move an existing customer from personal sign-in methods to enterprise SSO, with rollout choices, provider responsibilities, and cutover criteria.
- Decide whether permission inheritance needs explicit deny
Compare allow-only inheritance, deny precedence, and boundary objects through collaboration scenarios and operational tradeoffs.
- Should a suspended subscription prevent customer login?
A suspended subscription usually should not block authentication.
- Can a service account be renamed without changing its permissions or audit history?
Use immutable machine subject IDs for service accounts and treat names as labels. That preserves permissions and audit continuity through renames and safe name reuse.
- Should retention delete rows in batches or drop whole table partitions?
Choose partition drops when retention lines up cleanly with time boundaries; keep batch deletes when records need individual lifetimes, exceptions, or can outlive their cohort.
- What evidence should we capture before restarting a failing service?
A practical pre-restart evidence budget for on-call engineers: what to capture, when to stop, and how to retain transient failure state safely.
- Stop stale responses from overwriting newer UI state
Use cancellation, request identity, complete cache keys, and server-side conflict checks so older searches, validations, and navigations do not overwrite newer intent.
- OAuth redirect URI validation without wildcard surprises
Use exact registered redirects, constrained loopback handling, and explicit preview-environment patterns to prevent redirect and code-stealing flaws.
- What incident ownership should we agree on before adopting managed customer identity?
Define incident ownership before buying managed CIAM: who triages, what evidence crosses the handoff, and which failures stay with your app versus the identity provider.
- What should CI upload when a failed job cannot be reproduced locally?
A failed CI job should upload a small, predictable diagnostic bundle, not a giant copy of the workspace.
- Alternatives to routing customer login orchestration through an enterprise service bus
Should identity journeys remain in enterprise integration middleware or move to a product-owned application and identity boundary?
- How should a PATCH endpoint enforce different permissions for individual fields?
Treat PATCH as a set of explicit writes, authorize each requested field mutation, and fail closed on unknown or forbidden paths to prevent mass assignment.
- Should duplicating a project copy its members and sharing permissions?
Treat a duplicate as a new resource with a deliberate initial access policy.
- Should a public API sandbox use the same developer accounts as production?
Use one developer login across environments, but keep sandbox and production apps, secrets, tokens, data, and grants separate.
- Distributed rate limits need a consistency model too
Choose centralized, local, leased, or approximate quotas by fairness, latency, availability, and acceptable overshoot during failures.
- Why does database disk usage keep growing after we delete rows?
PostgreSQL usually reuses space freed by DELETE before it returns it to the operating system.
- Define null, missing, and empty before they cross a boundary
Give create, patch, clear, default, and unknown states field-specific meanings that survive JSON, forms, TypeScript, and SQL.
News
- Ai2 releases BenchMIRT to audit what benchmarks measure
BenchMIRT applies item response theory to model evaluations, helping researchers inspect which abilities benchmark questions actually distinguish.
- Anthropic previews customer-controlled Enterprise Frontier Safeguards
EFS separates activity-data custody from Anthropic's automated misuse detection, with customer-owned storage and review controls.
- Lambda adds full IAM resource-based policy support
Lambda expands resource policies to support multiple principals, actions, and IAM conditions, giving platform teams a broader policy surface to review.
- Sentry links OpenTelemetry spans to SDK-captured errors
Sentry’s OtlpIntegration connects existing OpenTelemetry trace context with Sentry errors across several backend SDKs, with explicit limits.
- Google launched Gemini Robotics ER 2 for robot planning and coordination
The embodied-reasoning model uses video and spatial context to plan tasks, monitor progress and coordinate more than one robot.
- GitHub Actions adds commit-pinned $/ references for local actions
The new self-repository syntax resolves actions and reusable workflows at the caller’s exact commit without a separate checkout step.
- Security Hub adds an organization-wide AI asset inventory
Security Hub combines managed-resource, software, and DNS signals to discover AI assets, giving security teams a new inventory to reconcile with owners.
- PostgreSQL 19 Beta 1 previews replica-wait and maintenance features
PostgreSQL 19's first beta introduces replica replay waits and maintenance changes, opening a testing window for application and driver compatibility.
- GitLab 19.0 adds secrets management and broader merge-request automation
The major release introduces a Secrets Manager beta, CI component analytics, and agent-assisted work across the merge-request lifecycle.
- AWS releases ExtendDB 0.1 with a PostgreSQL storage backend
ExtendDB brings the DynamoDB API model to pluggable storage, giving teams a new local and disconnected deployment option to test for compatibility.
- Microsoft expands MatterSim with a multi-task materials model
MatterSim’s May update combines a new multi-task model, simulation improvements and experimental follow-up on a predicted thermal conductor.
- Valkey Admin 1.0 unifies cluster inspection and troubleshooting
Valkey Admin 1.0 adds topology, key browsing and command logs, with server-version prerequisites and web-interface authentication to plan.
- Python 3.15 enters beta and asks maintainers to test compatibility
Python 3.15's first beta starts wider compatibility testing, with lazy imports and UTF-8 defaults among the changes package maintainers should examine.
- GPT-5.5 Instant becomes ChatGPT's default model
OpenAI replaces GPT-5.3 Instant with a more concise default and adds visible memory sources for personalized responses across consumer plans.
- Mistral puts Workflows into public preview
Mistral’s orchestration preview pairs a hosted control plane with customer-run workers, making failure recovery and approval handling key pilot questions.
- OpenAI launches GPT-5.5 for agentic coding and computer-based work
GPT-5.5 arrived in ChatGPT and Codex with stronger long-horizon tool use, a 400K Codex context window, and expanded cyber safeguards.
- Aurora Serverless platform version 4 changes scaling behavior
Aurora Serverless adds platform version 4 with scaling improvements, making workload transitions and platform-version tracking important upgrade checks.
- Firefox 149 ships hint popovers and math font support
Firefox 149 adds the hint popover state, math font defaults, and platform changes that affect layered overlays and mathematical content.
- Cloud SQL autoscaling read pools reach general availability
Enterprise Plus read pools can scale MySQL and PostgreSQL replicas behind one endpoint, but applications still need a clear read-consistency policy.
- Cloudflare introduces Custom Regions for customer-defined data boundaries
Custom Regions let enterprises define geographic processing boundaries, while Cloudflare expands its catalog of predefined regional controls.
- Valkey Search 1.2 adds text filters and server-side aggregations
Valkey Search 1.2 combines text, tags, numeric ranges and vectors; primary routing and shard boundaries matter when testing read-after-write behavior.
- GitHub releases REST API version 2026-03-10
A new dated REST API contract gives integrators a controlled migration point, with breaking changes isolated behind an explicit version header.
- Azure Compute API will always return a VM security type
A new Compute API version removes nullable securityType responses, requiring clients to treat Standard as an explicit value rather than an omission.