Terms - baki.io

What this site is

baki.io is a personal site - projects, experiments, essays, and machine-room documentation authored by Baki. Content is offered under terms that respect both author and reader.

Jurisdiction

baki.io is operated from Berlin, Germany. German law applies to these terms and to any use of this site, excluding its conflict-of-laws rules and excluding the UN Convention on Contracts for the International Sale of Goods (CISG).

Place of jurisdiction: Berlin, Germany. Where the visitor is a merchant (Kaufmann), a legal person under public law, or a special fund under public law within the meaning of §38 ZPO (Zivilprozessordnung), Berlin is the exclusive court of jurisdiction for all disputes arising out of or in connection with this site. Mandatory consumer-protection venues under EU Regulation 1215/2012 (Brussels I bis) are unaffected.

Copyright issues are governed by the German Urheberrechtsgesetz (UrhG). The DMCA (US-only) does not apply; the notice-and-takedown procedure below is the German-law equivalent.

Reading

Read freely. No account needed. No paywall. No tracking.

Content licensing

Unless otherwise noted on a specific page, original writing and imagery on this site is copyright Baki and licensed under Creative Commons BY-NC-SA 4.0. Code samples are MIT-licensed.

Notice-and-takedown

If you believe content on baki.io infringes your copyright under the Urheberrechtsgesetz (UrhG) — or any equivalent foreign right — you may submit a takedown notice. This is the German-law equivalent of the US DMCA process; the standard for a valid pre-litigation warning is §97a UrhG (Abmahnung).

How to file a notice. Send an email to [email protected] with the subject line Takedown notice — baki.io. To be actionable under §97a UrhG, the notice must contain:

Identification of the protected work — title, URL of the original, and a description of the work allegedly infringed.
Location of the alleged infringement on this site — the exact URL(s) on baki.io where the material appears, with anchor or screenshot if helpful.
Statement of authorization — confirmation that you are the rightsholder, an authorized agent, or hold an exclusive licence sufficient to assert the claim.
Good-faith statement — that the use is, to the best of your knowledge, not authorized by the rightsholder, an agent, or the law (including §51 UrhG quotation right and §44a–63a UrhG exceptions).
Contact details — full legal name, postal address (or that of your representative), and a return email address.
Signature — a typed name in the email body is treated as a qualifying signature for the purpose of initiating review.

What happens next. Notices are reviewed within a reasonable period (target: 7 calendar days). If the claim is facially valid, the disputed content is taken offline pending resolution and the uploader (where applicable) is notified. Frivolous or abusive notices may be rejected; bad-faith notices may themselves trigger liability under §97a(4) UrhG.

Counter-notice. If you posted content that was taken down and you believe the takedown was issued in error or covers material protected by an exception (quotation, parody, news reporting, scientific use), you may submit a counter-notice to the same email with the subject Counter-notice — baki.io. The counter-notice should mirror the structure above and add (a) identification of the removed material, (b) a statement under penalty of perjury (or its German-law equivalent: a declaration “an Eides statt”) that the takedown was a mistake or misidentification, and (c) consent to the venue specified in the Jurisdiction section. Restoration follows after a 10-business-day window during which the original claimant may seek interim relief.

Limit of role. baki.io is a single-author site, not a hosting provider in the sense of §10 TMG. The notice-and-takedown surface is offered as a courtesy; nothing here constitutes admission of host-provider status or expansion of statutory liability.

Signal drops (visitor comments)

When the signal-drop module is live, visitors may leave short contextual notes on content pages. By posting a signal drop you agree that:

You own what you post or have the right to post it.
No personal attacks, harassment, or illegal content.
Signal drops are moderated. The content safety pipeline may hide or remove any drop that violates these terms.
Signal drops may be displayed publicly and archived. You can request removal of your own drops at any time.

Processing of any data submitted with a signal drop (including draft state held in your browser) is described in /system/privacy. The moderation gate is part of the content safety pipeline referenced there; nothing visitor-submitted is published until it has cleared that gate.

Agent / scraper access

AI agents are welcome to read this site. The machine-readable metadata (RSS, structured data, the system collection, plus llms.txt and llms-full.txt) exists partly for you. When accessing the site, follow this policy:

Indexing crawlers (Google, Bing, DuckDuckGo, Brave Search, Kagi, and similar general-purpose search engines) — permitted. The full sitemap is exposed at /sitemap-index.xml.

LLM training and answer-engine crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Anthropic-ai, Applebot-Extended, and similar) — permitted, subject to the granular allowlist published in /llms.txt (the llmstxt.org convention). Surfaces marked private (such as /preferences and any /api/* endpoints) are off-limits even when not separately listed in robots.txt.

Misbehaving or undeclared crawlers — disallowed. Specific names are pinned in /robots.txt; the file is the authoritative machine-readable record.

When crawling:

Respect robots.txt and any Crawl-Delay directive declared there. Where no Crawl-Delay is set, treat one request per second as the courtesy ceiling per origin.
Identify yourself honestly in the User-Agent string. Disguised crawlers, residential-proxy reuse, or rotation through stealth IP pools to evade rate limits will be treated as undeclared and blocked.
Do not attempt to submit signal drops, write to admin endpoints, or interact with state-bearing UI (preferences, drop drafts, debate vertices, edit-mode controls).
Cite back when quoting. Attribution helps readers find the canonical source and keeps the licence terms in Content licensing legible.

Agents detected by behavioral signature are pinned to a deterministic output register (no ambient motion) to keep crawled content stable.

No warranty

This site is offered “as-is” (ohne jede Gewähr). Code shown here is illustrative, not production-hardened. Editorial content reflects opinion and personal experience, not professional advice.

To the extent permitted by German law, no warranty (Gewährleistung) is given as to fitness for any particular purpose, accuracy, completeness, or uninterrupted availability. The following carveouts apply and cannot be excluded under German consumer-protection and tort law:

Liability for intent (Vorsatz) and gross negligence (grobe Fahrlässigkeit) — not excluded.
Liability for injury to life, body, or health — not excluded (§309 No. 7(a) BGB).
Liability under the Produkthaftungsgesetz (ProdHaftG) where applicable — not excluded.
Liability for breach of material contractual obligations (Kardinalpflichten) — limited to typical foreseeable damage but not excluded.

Anything beyond those carveouts — incidental, consequential, or speculative loss arising from reading the site, copying code, or relying on essays — is at the reader’s own risk.

Changes

Version number and update date at the top of this page track revisions. Material changes will be called out in the changelog. Every commit-level snapshot is auto-archived to docs/legal/terms/ — see the revision index for the full chronology.