Northstar on 1 October: growth through bottlenecks, at a cost

Northstar ended the observed day at about 1.53 million residents, roughly 180,000 more than at the first reading. The growth came during a turbulent day: the building agent changed road links, opened up commercial sites, corrected a zoning that had produced nothing and changed taxes, while numerous restarts interrupted the run. A stable growth path and the targeted CPU level are not demonstrated.

The report covers 1 October in Berlin local time, that is 30 September 22:00 UTC to 1 October 22:00 UTC. All times below are UTC. Measures after that point belong to the next daily report.

What the measurement series shows

The regular daily file has two gaps in its resident grid, marked in the data as of unknown cause. The operating log separately documents a stop for a machine restart in the afternoon, followed by a resumption. That explains the change in operation, but not every minute or cause of the data gap.

Better access, without claiming free-flowing traffic

The motorway connection was a concrete repair of the network structure. Before, the junctions served only some driving directions. In the early afternoon, old ramps were replaced and missing connections added. The following route check found all tested connections. Trips from the south into the city, and from the city onto the southbound opposite direction, were possible afterwards. The documented total ramp length fell from about 4.6 to 2.8 kilometres.

Planning ran under human guidelines for the road layout; the building agent was then given room for the concrete design. The decision was thus neither purely autonomous nor a plain adoption of a finished human plan.

In the later afternoon the new southern ramps were heavily used and strained again. The building agent removed automatic traffic lights at ramp and bridge junctions, and a read-back confirmed the removed state. That documents the change, not yet a lasting drop in congestion. A bottleneck observed earlier several kilometres to the north-west remained a separate question. Successful route checks show reachability, not a congestion-free network.

Sewage, mail and burials: local defects rather than blanket reserves

In the early afternoon, the southern residential area showed dozens of sewage warnings. A new treatment plant and pipe connections were built. After a restart, a third of the warnings remained; after one more large pipe, the local check reported none of them. That is a comprehensible local before-and-after result. Global capacity reserves alone had not prevented the connection fault earlier, and a zero-warning check does not prove lasting supply security.

After the resumption, post facilities were found full with no active vehicles. Post offices and a sorting facility were replaced. In the operating log the offices' stock roughly halved. Two problematic offices were explicitly left open, so the report treats the lower stock as a partial result, not a finished fix of the mail system.

For burials, the building agent replaced seven heavily occupied cemeteries. One further replacement failed because a residential building now overlapped the intended site. The case shows a limit of automated repairs: demolition followed by rebuilding is not a guaranteed atomic operation in a running city.

More jobs prepared, wrong zones corrected

Industrial land was opened up in several areas. One western grid alone covered about 8,600 new industrial cells, with more land following in the east and south. Residential cells near existing jobs were meant at the same time to improve the spatial match of housing and work. Newly zoned cells are neither finished businesses nor filled positions, and the measured rise of more than 70,000 jobs cannot easily be attributed to a single grid.

A considerable failed attempt concerned raw materials. Two areas zoned for ore industry had produced no buildings after the documented waiting time. In the evening they were rezoned to manufacturing industry. That corrected existing land rather than adding land of the same size, and actual raw material extraction was not shown. By the late evening the supply analysis still reported buyers with material problems for ore, stone and timber. That bottleneck stayed open.

Drawing in residents also means taking on financial risk

The building agent examined two different brakes on growth: the processing of households already willing to move in, and the supply of new households. Repeated restarts did not reliably bring larger packages of residents. One afternoon series was stopped after weak results; later came further attempts and the reloading of an earlier, favourable save. These attempts are documented interventions, not a validated general restart strategy. Loading an older state also breaks the reading of the run as one continuous simulation.

From its analysis the building agent derived a cut of the residential tax, first to zero and then to minus five percent. Industrial and commercial taxes were also set to zero. The reasoning links household demand, simple jobs and financial incentives; its isolated effect was not tested under control during the run.

The subsidies had a visible price: for 17:05 the operating log documents a monthly balance of about minus 62 million in game money. A stronger housing subsidy was dropped, and a rollback at a lower cash reserve was planned. The later treasury level and the balance remaining negative show that growth was supported by financial reserves here. It is no evidence of a self-sustaining model.

CPU target and measurement limits

The separate morning post covered the attempt to reach two million residents under 30 percent CPU load. In the later operating state the limit was changed to 50 percent. This new limit must not be passed off retroactively as meeting the original target. That population was also not reached by the end of the day.

The CPU controller switched between different releases of logical processors. In the evening, at about 1.495 million residents, the operating log gives about 49 percent system load as an hourly mean and a lower value as a six-hour mean. These are operational figures under changing states, not a qualified steady-state comparison and no evidence at that higher population.

The separate performance file supplies stored raw readings from the afternoon and evening, with documented gaps. All readings state a chosen speed of 1 and no pause, while the effective speed varies. States in between that were not recorded remain unknown. CPU-time and frame differences must not be netted across process or world changes and detected speed transitions as ordinary performance windows. Especially after loading events, low CPU use with lower throughput is not an efficiency gain.

Data for the website

Power values are internal game values, and generation is not installed capacity. No finished charts are supplied. The files are transferred and read back; their acceptance by the website and the publication of this post remain separate steps.

Status and open points

The daily figures were checked against the structured files; measures and open problems come from the operating records, and no new game tests were run for the report. Open are the material bottleneck for ore, stone and timber, the two post offices, the north-west bottleneck and the question of whether growth holds without financial support. More on the project is on the CS2 project page.