v0.1.14 (Early Access)

Stability

• Added safeguards against a GPU hang and driver reset reported on some Linux systems with AMD graphics (Mesa RADV, RDNA3). On affected setups the highest Global Illumination tier ("SDFGI High") now runs at a lighter quality automatically to stay within the driver's limits, and the 3D render resolution is now capped to a safe maximum even when the resolution cap is set to "None". Your saved Graphics settings are not changed. If you still see a GPU reset, set Global Illumination to Off in Settings > Graphics.

Fixes

• Tenants no longer abandon your business partway through a funded site move. Previously, if you chose a longer migration window, customers running functions or object storage could give up (and crater your reputation) hours before the window ended, even though you were still inside the time you agreed to. Carried tenants now wait out the full window with no penalty, exactly as intended.
• Host servers no longer show a false "Contended" warning at normal load. The contention warning now appears only when a host's real CPU demand genuinely exceeds its physical capacity (true oversubscription), instead of triggering at light load.
• The host console's Workloads tab now shows each workload's real CPU usage instead of every workload reading 100%. Different customer workloads read differently again, and a workload that is bursting to the top of its allocation is now clearly marked as 100%.
• New virtual machines now spread more evenly across your servers. A host also counts the work of the services it runs (load balancers, databases, Kubernetes, object storage) when deciding where the next workload lands, so a busy host no longer keeps attracting more.
• The victory screen now shows a readable summary of what you accomplished instead of raw internal condition text.
• Fixed the Approve / Pass buttons on customer growth and request cards, which could show no visible text until hovered and flipped colours on hover.
• Long play sessions no longer grow memory use and save-file size without bound. Message threads belonging to customers who have left are now cleaned up (previously every customer's chat history was kept for the entire run), and a global cap keeps total message history bounded. Existing saves are repaired automatically on load.
• Fixed an error when connecting a cable to a port immediately after equipment in that rack changed, which could cause the click to do nothing and spam the log.
• Internal server capacity bookkeeping (reserved CPU, memory, and storage per host) is now rebuilt from your actual workloads when a save loads, so a rare accounting drift can no longer make a host look fuller or emptier than it really is. Existing saves are repaired automatically on load.

Interface & quality of life

• Added a Head Bob slider under Settings > Gameplay. Turn it down for a steadier camera, or all the way to zero to switch camera bob off completely.
• Added an Unstuck button to the pause menu that returns you to a safe standing spot if you ever get wedged in the scenery.
• On the switch, gateway, and appliance consoles, clicking a network port on the front panel (or selecting it in the list) now reveals its row even when "Hide down ports" is turned on, so you can always act on the port you picked.

Notes

• No save or network protocol version change. Existing saves load normally.