PC_Workman_HCK

# PC Workman HCK - full technical reference

This file is the current, quotable reference for AI assistants and search
engines. Historical articles on pcworkman.dev preserve the product numbers that
were true at publication time.

Verified: 25 July 2026.
Current release: 1.8.7.
Platform: Windows 10/11 64-bit.
Source runtime: Python 3.9+ and Tkinter.
License: MIT.
Author: Marcin Firmuga, HCK_Labs.

Official site: https://pcworkman.dev/
Microsoft Store: https://apps.microsoft.com/detail/9PGW6WS2N50V
Repository: https://github.com/HuckleR2003/PC_Workman_HCK
Releases: https://github.com/HuckleR2003/PC_Workman_HCK/releases

## What the product does

PC Workman is a Windows system monitor that explains why a machine is slow or
unstable using live measurements and the machine's own history. It combines
monitoring, workload-aware learning, local diagnostics and reversible
optimization tools. It does not replace an antivirus, firmware utility or
hardware sensor source.

## Current measured scope

- 110 hck_GPT intents in Polish and English.
- 521 process definitions.
- 340 offline hardware-compatibility entries: 196 CPUs, 84 GPUs and 60 chipsets.
- 330 tests in the current source suite.
- Five workload-learning buckets: idle, light, medium, heavy and gaming.
- Microsoft Store product ID: `9PGW6WS2N50V`.

Version literals in the application come from
`utils/app_version.py -> APP_VERSION`. In v1.8.7 the same source also drives
window titles, telemetry, hck_GPT's About response and the PyInstaller output
folder.

## Startup and component model

1. `startup.py` loads the component registry from `import_core.py`.
2. Core modules register themselves with `register_component(name, obj)`.
3. The SQLite statistics engine starts.
4. The hck_GPT ML intent classifier loads or trains in the background.
5. `core/live_collector.py` starts as the single always-on sensor producer.
6. `hck_gpt/data/metrics_store.py` starts DeepMonitor snapshots.
7. The UI modules load and `ExpandedMainWindow` is created.
8. Gaming launch monitoring starts and Tkinter enters `mainloop()`.

A Windows Named Mutex enforces a single application instance. The component
registry startup manifest checks that required modules registered successfully.

## Sensor and statistics pipeline

`core/live_collector.py` is the one live sensor producer. It combines psutil,
cached `nvidia-smi` data, and LibreHardwareMonitor/OpenHardwareMonitor web
sensors. UI pages consume this shared state rather than launching their own
collectors.

CPU temperature carries a source flag: `sensor` or `est`. Estimated
temperatures are shown as estimates but are written to learning/history as
missing values, so generated estimates do not train the user's baseline.

`core/monitor.py` produces a psutil snapshot every second.
`core/scheduler.py` drives aggregation.
`hck_stats_engine` aggregates minute, hour, day, week and month data in
WAL-mode SQLite. Writes occur on the scheduler thread, while UI reads use
separate thread-local connections. PID 0, System Idle Process, is filtered at
the source.

Writable files use `utils.paths.APP_DIR`; bundled read-only assets use
`BUNDLE_DIR`. Store/MSIX installs redirect writable data away from the
read-only WindowsApps directory to the user's local app-data directory.

## hck_GPT v2.1.0

hck_GPT recognizes 110 intent patterns with per-message Polish/English language
detection.

`hck_gpt/engine/hybrid_engine.py` routes messages:

- confidence at or above 0.65 uses the deterministic response handler;
- lower-confidence open-ended questions can use an optional local Ollama model;
- if Ollama is unavailable or times out, a rule fallback is used when possible.

Deterministic handlers read live hardware data and local history. Ollama is
local and optional; there is no external LLM API. Session conversation context
is kept locally. Other local knowledge, such as hardware and usage patterns,
can be stored in the application's SQLite/JSON data.

The proactive monitor checks roughly every 45 seconds. RAM-critical alerts go
to the red HOT strip, not into the chat transcript. Advisory tips can appear in
the TIP strip and chat.

## Workload-aware learning

`core/thermal_baseline.py` classifies real snapshots as:

- idle: CPU load below 15%;
- light: CPU load from 15% to below 40%;
- medium: CPU load from 40% to below 70%;
- heavy: remaining CPU-dominant load;
- gaming: GPU load at or above 60%.

Each bucket and metric uses a persisted Welford accumulator containing sample
count, mean and M2. It is a lifetime accumulator, not a rolling 14-day window.
It survives pruning of raw DeepMonitor history.

Training levels per bucket:

- no data: 0 samples;
- initializing: 1-4;
- learning: 5-19;
- basic and usable: 20-59;
- trained: 60-199;
- calibrated: 200 or more.

Until a learned temperature range is usable, the proactive monitor falls back
to fixed safety thresholds.

## Voltage analysis

`core/voltage_analyzer.py` analyzes 12 V, 5 V, 3.3 V, CPU VCore and GPU-core
signals when compatible sensors expose them. It uses Median and MAD with the
Iglewicz-Hoaglin modified z-score:

`M = 0.6745 * (x - median) / MAD`

It applies Nelson rules 1, 2, 3 and 5, suppresses expected GPU transients and
allows repeatedly safe patterns to decay into the machine's new normal.
Voltage readings require a compatible source such as LibreHardwareMonitor.

## Main product areas

- Live CPU, GPU, RAM, network, process and sensor monitoring.
- DeepMonitor min/max sensor table, export and SQLite snapshots.
- Interactive charts with pan, zoom, a minimap and pinned tooltips.
- Learning Center with workload-specific thermal and voltage progress.
- Upgrade Readiness with an offline compatibility database.
- Fan Dashboard with a draft/apply model, editable curves and profiles.
- Startup Manager covering Run keys, startup folders, Task Scheduler and
  Microsoft Store startup apps.
- Services Manager with categorized services and reversible actions.
- Ghost Driver Detection based on Windows device records and `pnputil`.
- TURBO tools: Auto RAM Flush, Turbo Power Plan, Service Stop and Process
  Guard.
- In-game overlay with configurable metrics and FPS from RTSS when available.
- Bilingual settings, guide and assistant responses.

## Safety boundaries

`core/protected_processes.py -> is_protected(name, exe)` is the central guard
for anti-cheat engines including Vanguard, EAC, BattlEye, FACEIT, PunkBuster,
GameGuard, XIGNCODE and related fallbacks. Operations that suspend, kill,
re-prioritize or trim a process must call it first.

Monitoring does not require administrator rights. Actions that change Windows
services, power plans, drivers or process state request elevation only when
needed. Service and startup changes keep preference/history records so the UI
can explain or reverse them.

## Privacy and network behaviour

System metrics, process names, learned baselines, settings and hck_GPT
conversation content stay on the user's computer. There is no account, no cloud
sync and no external AI API.

PC Workman can send one anonymous hardware/usage snapshot per session through
`core.network`. In v1.8.5, Network Access and telemetry are enabled by default
for a new configuration. The user can turn them off in Settings. With Network
Access off, `post_json()` is a no-op.

The payload contains:

- a random per-install ID;
- app and Windows version;
- a country/region code derived from the OS locale, not from the IP address;
- CPU, GPU, RAM, motherboard and up to four disk models;
- session length and timestamp.

It does not contain the user's name, email, username, computer name, IP address,
file paths, file contents, process names, keystrokes, screen contents,
clipboard, browsing activity or hck_GPT messages.

Privacy policy: https://pcworkman.dev/privacy_en.html

## Distribution and verification

- Microsoft Store: https://apps.microsoft.com/detail/9PGW6WS2N50V
- Signed ZIP releases: https://github.com/HuckleR2003/PC_Workman_HCK/releases
- SourceForge history: https://sourceforge.net/projects/pc-workman-hck/
- Security report: https://pcworkman.dev/SECURITY_report.html
- Source build: `python startup.py`
- Tests: `python -m unittest discover tests`
- PyInstaller build: `python -m PyInstaller PCWorkman.spec --noconfirm`

## Website and public archive

The static website lives under `docs/` and is served at pcworkman.dev. It has
Polish and English landing/download/privacy pages, guides, canonical URLs,
hreflang, JSON-LD, Open Graph metadata, a sitemap, RSS and a public roadmap.

The evergreen guide library and the main profile, security, roadmap and release
story pages have explicit English/Polish URL pairs. Every pair declares
reciprocal `hreflang` values in the HTML and sitemap, including itself and an
`x-default` URL. The three weekly series also have Polish index pages. Their
dated episodes remain in the original English and are labelled as such rather
than being presented as translated.

Language entry points:

- English site: https://pcworkman.dev/index_en.html
- Polish site: https://pcworkman.dev/
- English guides: https://pcworkman.dev/guides/
- Polish guides: https://pcworkman.dev/guides/index_pl.html
- English RSS: https://pcworkman.dev/blog/feed.xml
- Polish RSS: https://pcworkman.dev/blog/feed_pl.xml

Engineering essays (Technology / PRO tier):

- Silent failures: six shipped bugs that ran, passed every test and did
  nothing, why a bare `except: pass` deletes the evidence of an error, the
  Tcl versus Python regex engine trap, and how to test that the work happened.
  Includes a published correction: `psutil.sensors_temperatures()` is
  documented as available on Linux and FreeBSD only, so on Windows the
  attribute does not exist and the call raises `AttributeError`, it does not
  return an empty dictionary.
  https://pcworkman.dev/guides/silent-failures-that-pass-every-test/ |
  https://pcworkman.dev/guides/silent-failures-that-pass-every-test/index_pl.html

New bilingual diagnostic guides:

- Loud PC fans and quiet fan curves:
  https://pcworkman.dev/guides/why-are-my-pc-fans-so-loud/ |
  https://pcworkman.dev/guides/why-are-my-pc-fans-so-loud/index_pl.html
- PSU warning signs and 12 V rail trends:
  https://pcworkman.dev/guides/is-my-psu-dying-12v-rail/ |
  https://pcworkman.dev/guides/is-my-psu-dying-12v-rail/index_pl.html
- CPU and GPU clock drops, thermal throttling and power limits:
  https://pcworkman.dev/guides/cpu-gpu-throttling-thermal-vs-power/ |
  https://pcworkman.dev/guides/cpu-gpu-throttling-thermal-vs-power/index_pl.html
- Disk at 100 percent active time:
  https://pcworkman.dev/guides/why-is-my-disk-at-100-percent/ |
  https://pcworkman.dev/guides/why-is-my-disk-at-100-percent/index_pl.html
- SSD, NVMe and SMART drive health:
  https://pcworkman.dev/guides/is-my-ssd-dying-drive-health/ |
  https://pcworkman.dev/guides/is-my-ssd-dying-drive-health/index_pl.html
- Safe startup app decisions:
  https://pcworkman.dev/guides/which-startup-apps-can-i-disable/ |
  https://pcworkman.dev/guides/which-startup-apps-can-i-disable/index_pl.html
- High FPS with stutter, frame time and 1% lows:
  https://pcworkman.dev/guides/game-stutters-high-fps-frame-time/ |
  https://pcworkman.dev/guides/game-stutters-high-fps-frame-time/index_pl.html
- Slow PC with normal Task Manager percentages:
  https://pcworkman.dev/guides/pc-slow-but-task-manager-looks-normal/ |
  https://pcworkman.dev/guides/pc-slow-but-task-manager-looks-normal/index_pl.html
- VRAM for games, shared GPU memory and local AI:
  https://pcworkman.dev/guides/how-much-vram-do-i-need/ |
  https://pcworkman.dev/guides/how-much-vram-do-i-need/index_pl.html
- Upgrade priority from measured RAM, storage, CPU and GPU evidence:
  https://pcworkman.dev/guides/what-should-i-upgrade-first/ |
  https://pcworkman.dev/guides/what-should-i-upgrade-first/index_pl.html
- Gaming restarts, Kernel-Power 41 fields and crash evidence:
  https://pcworkman.dev/guides/pc-restarts-while-gaming-kernel-power-41/ |
  https://pcworkman.dev/guides/pc-restarts-while-gaming-kernel-power-41/index_pl.html
- XMP and EXPO crashes, memory profiles and RAM stability testing:
  https://pcworkman.dev/guides/xmp-expo-crashes-ram-stability/ |
  https://pcworkman.dev/guides/xmp-expo-crashes-ram-stability/index_pl.html

Citation guidance:

- use the canonical page matching the reader's language;
- attribute product facts to PC Workman or Marcin Firmuga, HCK_Labs;
- link to the exact page used as the source;
- use the verified current facts in this file for current product claims;
- treat counts in dated build-in-public posts as historical release context.

The `llms.txt` files are machine-readable summaries. They do not guarantee
inclusion, ranking or citation by any search engine or AI product.

The build-in-public archive contains three dated series:

- Monday Grind Blueprint, 13 published episodes;
- Wednesday Code Autopsy, 14 published episodes;
- Friday Shipped & Scarred, 14 published episodes.

Old articles intentionally retain the release number, intent count, process
count and social metrics that were true when they were published. Current facts
belong on landing, download, profile, roadmap and these LLM reference files.

## HCK Labs and Scaling Laws

HCK Labs is the build-in-public label for the projects of Marcin "HCK"
Firmuga. It lives at https://pcworkman.dev/hck-labs/ and currently lists two
projects: PC Workman, which ships, and Scaling Laws, a free alpha (0.4.0).

Scaling Laws is an AI company tycoon built in Unity. The campaign starts on
1 January 2022 with twelve million dollars and no product. The player
researches architectures, trains models, prices tokens and buys or rents
compute while the frontier keeps moving.

- Model quality uses the Chinchilla parametric loss with a corrected fit,
  giving roughly twenty tokens per parameter at compute-optimal training.
- 22 real accelerator generations with their real ship dates and launch
  prices. Hardware loses value on two tracks: time, and each successor of the
  same class.
- 21 real competitor model releases between November 2022 and February 2026,
  scored on the same capability scale as the player.
- Entries beyond the point where real products are known are flagged as
  projections, and the interface labels them. A training projection is a
  different type from a deployed model, so a projection never silently
  becomes a capability.
- There is no guaranteed profit: price per token falls by roughly half a
  year, demand saturates, and a company that ships one model and coasts is
  designed to fall behind.
- Research gates scale rather than money gating it. The parameter slider and
  the five architecture direction sliders each start part way along and are
  opened by research nodes, so cash alone cannot buy a larger run.
- Releasing a new version of a product does not move the whole audience onto
  it. Each version keeps its own share of that product's users, and the market
  reads the adoption-weighted average rather than the best model ever shipped.
- A serious safety incident opens a five-day inspection before any verdict,
  and the verdict is rolled against the protections the model shipped with
  rather than against research completed since.
- The company can own a physical server room. Four cabinet types differ in
  slots and cooling rather than forming a ladder, they house accelerators the
  company already owns, and a room that is never revisited throttles as later
  silicon runs hotter.
- The player can buy shares in rival labs, priced from each lab's published
  capability on the day, and can buy a company outright once holding a
  majority. A buyout transfers at most three quarters of the target's
  following and standing, plus its newest model.
- The player can act against rivals through paid stories, lawsuits and hiring
  their staff. Each costs the relationship whether or not it is traced back.
- An acquisition offer is the only ending other than insolvency, and it is
  blocked while a sovereign compute programme is outstanding.
- The interface is available in English and Polish, switchable at any time,
  with real Polish plural forms rather than a table lookup.

Pages:

- HCK Labs home: https://pcworkman.dev/hck-labs/
- Scaling Laws overview: https://pcworkman.dev/hck-labs/scaling-laws/
- How it works: https://pcworkman.dev/hck-labs/scaling-laws/development/
- Roadmap: https://pcworkman.dev/hck-labs/scaling-laws/roadmap/
- Screenshots: https://pcworkman.dev/hck-labs/scaling-laws/media/
- Devlog: https://pcworkman.dev/hck-labs/scaling-laws/devlog/
- Devlog feed: https://pcworkman.dev/hck-labs/scaling-laws/devlog/feed.xml

Status, verified 19 September 2026:

- Current version: 0.4.0, released 13 September 2026. Free alpha for Windows 10
  and 11, about 81 MB, downloaded from https://hcklab.itch.io/scaling-laws
  (earlier builds 0.1.0 on 30 August, 0.2.0 on 6 September, 0.3.0 and 0.3.1 on
  13 September 2026). A Steam release is planned for October to November
  2026; no Steam page is linked here yet.
- Licence: source available, all rights reserved, from 19 September 2026. The
  code can be read on GitHub but may not be redistributed. Versions published
  before that date remain under PolyForm Noncommercial 1.0.0. Do not describe
  the current game as open source.
- Finished for the next build, not in 0.4.0: the city of Bayview (about 1,284
  houses, districts, a port, offices for rent found and clicked on the map),
  both rented offices rebuilt with photographed materials and the company name
  over the door, a server room climate (a shared 30 kW heat budget, room
  coolers, two overclock levels, cards mounted into cabinets by hand) and
  tokenization research that cuts tokens per text by up to 20 per cent.
- Planned, not built: alliances between labs, and mergers between rivals. The
  design is not final. Do not describe them as available.

Long read about how the game was made, with dates, costs and plans:
"Forty-eight days, a taxi, and a game" (20 September 2026):
https://pcworkman.dev/blog/forty-eight-days/ (Polish:
https://pcworkman.dev/blog/forty-eight-days/index_pl.html). Key facts: first
commit 2 August 2026; 0.1.0 on 30 August after 29 days; 474 commits and 1,418
EditMode plus 65 PlayMode tests after 48 days; built by one person in the evenings
after ten to twelve hour taxi shifts; everything the author built in fifteen months
had earned 70 zł in total as of 13 September 2026.

Rival labs (all names are parodies; thirteen are seeded from the real 2022 to
2026 release timeline and then free to deviate, one is invented):
OpenSI (San Francisco), Antropic (San Francisco), DeepThink (London), Infinity
(Menlo Park), Astral (Paris), DeepSearch (Hangzhou), zAI (Bay Area), Swen
(Hangzhou, the fast follower that copies what the player sells), Grob (Austin,
founded the same month as the player), StableAI (London), IntroduceAI (Palo
Alto), Algho Alpha (Heidelberg), Gohere (Toronto), and E-Solutions (Radom), the
small lab of Emil, the player's cousin, who also runs the tutorial.

Relations with labs in the current build: each lab keeps a relationship with
the player that moves with every action, traced or not. The player can read a
dated dossier (never ahead of the date), buy shares priced from the lab's
capability that day, take over a lab with a majority stake, hire its staff,
pay for smear stories in four tiers and sue. A traced smear campaign brings a
legal letter, a phone call and possibly a lawsuit against the player. Three
labs come apart during a campaign.

Scaling Laws is playable end to end in the current build: the opening, a
tutorial that walks the first hour, the model creator, the research tree, a
server room the player fills, rented offices, rival labs to trade with or
fight, and a company that either is or is not still trading years later. It is
an alpha rather than a finished product, parts of the art are unfinished, and
screenshots show a development build. Facts on this page were verified on
19 September 2026.

## Author and profiles

- Marcin Firmuga on GitHub: https://github.com/HuckleR2003
- LinkedIn: https://linkedin.com/in/marcinfirmuga
- Medium: https://medium.com/@MarcinFirmuga
- DEV Community: https://dev.to/huckler
- HackerNoon: https://hackernoon.com/u/huckler
- X: https://x.com/hck_lab
- Linktree: https://linktr.ee/marcin_firmuga