Replit
Each of these is a real decision Replit is working through, written so you can make the call yourself — incomplete information, someone who pushes back, no answer key.
Replit is a browser-based software development and deployment platform that unifies a cloud coding workspace, AI-assisted app building, collaboration, hosting, and built-in application services. It serves individual developers, learners, entrepreneurs, product teams, nontechnical business users, and enterprises building internal tools, prototypes, and production applications. Companies like Visa, Ernst & Young and Comcast are rolling Replit out to thousands of employees so finance, operations and product teams can build without waiting on an engineering backlog. Replit builds single-tenant and dedicated deployment options so customers can run Replit with the isolation guarantees their security posture requires. Replit runs large-scale cloud deployments with a deep focus on Kubernetes, Docker, and GCP. Its systems use TypeScript, Go, Rust, Postgres, gRPC/Protobuf, Kubernetes, Envoy, and Restate.
Your first one is free. No card.
31 live right now.
The rest
Build a tamper-evident record of Agent actions until admins widen access
Administrators cannot consistently connect an Agent action to the user, permission, policy decision, and workload behind it. You have to weigh audit-ready evidence against the friction that slows approved Agent work.
Contain audit access before the evidence disappears
Two enterprise administrators found audit entries for apps their groups cannot access. You must weigh immediate credential containment against preserving enough evidence to establish scope.
Refresh workspace access after SCIM changes without widening permissions
Approved builders remain locked out after their group membership updates. You must weigh faster entitlement refresh against fail-closed access control.
Move a known failure to engineer approval without unapproved changes
Recurring deployment failures leave builders retrying publishes while on-call engineers reconstruct familiar context. You must weigh faster runbook execution against evidence an engineer can trust before production changes.
Trace failed deployments to ranked causes so engineers confirm faster
Private-connectivity investigations are delaying enterprise deployments because engineers cannot quickly confirm the cause. You have to weigh faster evidence assembly against the review needed before anyone changes a customer environment.
Route workload alerts while responders reconstruct agent context
Suspected workload-isolation alerts force responders to piece together cloud and Kubernetes evidence while the incident is active. You must weigh faster severity decisions against the risk of routing on unverified inference.
Prove builder value before finance questions the next credit purchase
Administrators have builder activity but no agreed evidence that the spend produced a useful internal application. You must weigh a repeatable value path against the different governance and outcome needs of each pilot organization.
Reconcile AI usage before a regulated app moves to production
A regulated internal application is paused because finance cannot reconcile the AI usage behind its expected charges. You must weigh a fast production decision against evidence that security and finance can both defend.
Close the SIEM gap for publishing events without loosening permissions
Publishing events appear in Audit Logs but fail to reach the customer’s SIEM. You have to restore the evidence trail without loosening the permissions their IT administrator protects.
Surface renewal risk before CSMs discover it on the call
Critical account signals are scattered across HubSpot, Zendesk, Linear, and Slack until a renewal conversation is already being prepared. You must weigh earlier ranking against the risk of letting weak evidence shape a customer conversation.
Trace publish failures while builders wait to launch
Publishing failures leave entrepreneurs retrying changes without knowing what actually broke. You must weigh faster investigation against the evidence engineers need before shipping a fix.
Turn launch briefs into approved drafts before buyer intent moves
Launch briefs are reaching review without agreement on the prompts and claims that matter to nontechnical builders. You have to weigh faster campaign preparation against evidence checks and strategist control of the final message.
Win 300 first deployments from spreadsheet replacers within six weeks
Builders looking to replace spreadsheets are leaving before they start a project. You have to weigh a focused first-use-case motion against the broader range of apps they want to build.
Set the credit story before launch materials lock
Comparison visitors are leaving before qualified project starts, and credit objections are surfacing in field conversations. You must weigh a simpler pricing story against claims that remain credible for different app requirements.
Fix the credit explanation so visitors start an Agent project
Qualified visitors are reaching the comparison page but leaving at the credit explanation. You must weigh a simpler promise against the accuracy needed for variable task costs.
Choose speed or flexibility for the first path before the pilot starts
Dedicated deployment requests currently require 8 to 12 manual infrastructure handoffs before an application can ship. You must weigh a fast first self-service path against an interface that remains safe and supportable after the pilot.
Commit the regional cutover before maintenance closes in
1,200 stateful workloads must move within 72 hours, while 9% of generated dependency maps conflict with sampled traces. You must weigh continuity under a hard deadline against the evidence required to protect isolation and recoverability.
Cut the publish p95 from 240 seconds to 90 while builds burst
Publish p95 has reached 240 seconds as shared Kubernetes nodes saturate during build bursts. You must weigh faster publish capacity against stable deployed workloads and unnecessary cloud spend.
Give founders an operating rhythm after their first publish
Premium founders stop publishing after small production concerns become urgent support cases. You must weigh a useful operating programme against the risk of becoming each customer’s engineering team.
Settle a renewal path in nine days without pausing active builders
A Premium account with 180 active builders may restrict access nine days before renewal because credit charges are not predictable. You have to weigh builder adoption against a commitment procurement can trust.
Restore apps for Premium builders without sending unverified rollbacks
Premium builders are republishing broken Agent revisions and reopening the same support cases. You must weigh fast recovery guidance against the risk of sending an unverified rollback path.
Show task costs before builders spend through their monthly credits
Core builders abandon revisions after unexpected credit use during substantial Agent tasks. You have to weigh useful cost guidance against estimates that overstate certainty.
Choose a schema migration before founders publish into the split
Published apps are reading older schemas, and founders are seeing live features fail after deployment. You have to weigh rapid recovery against irreversible changes to production data.
Restore custom domains while founders wait to launch
Custom-domain activation succeeds in only 81% of newly published apps, leaving founders with unreachable public URLs. You have to weigh a narrow publishing-path correction against the risk of changing existing domain mappings.
Put guardrails on Agent use without stalling internal builders
Administrators can approve models, yet group-level usage remains hard to control before costs arrive. You must weigh firm limits against a clear, productive path for employees using approved models.
Fix key rotation so no records go unreadable before cutover
A customer key rotation leaves some encrypted application records unreadable just before a fixed cutover. You must weigh a recoverable launch path against the encryption guarantees the customer will not relax.
Restore SSO access while keeping workspace permissions tight
Valid workspace members are looping through sign-in despite appearing in the right SSO groups. You have to restore access while preserving the permission boundary administrators rely on.
Hold agent actions on live data for approval without slowing builders
Entrepreneurs lack a customer-facing checkpoint before agents take sensitive production actions. You have to weigh meaningful approval evidence against the speed that makes natural-language application building useful.
Contain the tampering in the tool path without halting live apps
Eleven tool calls diverged from approved intent before unusual PostgreSQL audit activity appeared. You have to weigh immediate global containment against interrupting legitimate live applications.
Fix denied reads for new apps without widening workspace access
Denied tool calls rose to 6.8% in two hours for newly published applications. You have to restore legitimate access without weakening the boundary between workspaces.