docs: fix broken links found in audit

- update Arbol main link to arbol.io (company rebranded)
- replace broken gateway.arbolmarket.com with archive.org snapshot
- replace broken waterlily.ai with archive.org snapshot

Author: lidel

docs/case-studies/arbol.md

@@ -17,7 +17,7 @@ _— Ben Andre, CTO, Arbol_
 <img src="./images/logo-arbol.svg" alt="Arbol logo" width="220">
 :::
 
-[Arbol](https://www.arbolmarket.com/) is a software platform that connects agricultural entities like farmers and other weather-dependent parties with investors and other capital providers to insure and protect against weather-related risks. Arbol's platform sells contracts for parametric weather protection agreements in a marketplace that's an innovative, data-driven approach to risk management, cutting out the usual legacy insurance claims process of making loss assessments on the ground. Instead, Arbol relies on tamper-proof data indexes to determine payouts, and doesn't require a defined loss to be indemnified. Arbol's platform combines parametric weather protection with blockchain-based smart contracts to provide cost-efficient, automated, and user-defined weather-related risk hedging. As with traditional crop insurance and similar legacy products, end users purchase assurance that they'll be financially protected in the case of adverse weather — but with Arbol, these end users are paid automatically if adverse conditions occur, as defined by the contract and measured by local meteorological observations tracked by Arbol's data sources.
+[Arbol](https://www.arbol.io/) is a software platform that connects agricultural entities like farmers and other weather-dependent parties with investors and other capital providers to insure and protect against weather-related risks. Arbol's platform sells contracts for parametric weather protection agreements in a marketplace that's an innovative, data-driven approach to risk management, cutting out the usual legacy insurance claims process of making loss assessments on the ground. Instead, Arbol relies on tamper-proof data indexes to determine payouts, and doesn't require a defined loss to be indemnified. Arbol's platform combines parametric weather protection with blockchain-based smart contracts to provide cost-efficient, automated, and user-defined weather-related risk hedging. As with traditional crop insurance and similar legacy products, end users purchase assurance that they'll be financially protected in the case of adverse weather — but with Arbol, these end users are paid automatically if adverse conditions occur, as defined by the contract and measured by local meteorological observations tracked by Arbol's data sources.
 
 To build the data indexes that Arbol uses to handle its contracts, the team aggregates and standardizes billions of data files comprising decades of weather information from a wide range of reputable sources — all of which is stored on IPFS. IPFS is critical to Arbol's service model due to the inherent verifiability provided by its [content-addressed architecture](../concepts/content-addressing.md), as well as a decentralized data delivery model that facilitates Arbol's day-to-day aggregation, synchronization, and distribution of massive amounts of data.
 
@@ -88,7 +88,7 @@ Arbol's end users enjoy the "it just works" benefits of parametric protection, b
 
 8. **Pinning and syncing:** When storage nodes in the Arbol network detect that a new hash has been added to the heads file, they run the standard, recursive [`ipfs pin -r`](../reference/kubo/cli.md#ipfs-pin) command on it. Arbol's primary active nodes don't need to be large in number: The network includes a single [gateway node](../concepts/ipfs-gateway.md) that bootstraps with all the parsing/hashing nodes, and a few large storage nodes that serve as the primary data storage backup. However, data is also regularly synced with "cold nodes" — archival storage nodes that are mostly kept offline — as well as on individual IPFS nodes on Arbol's developers' and agronomists' personal computers.
 
-9. **Garbage collection:** Some older Arbol datasets require [garbage collection](../concepts/glossary.md#garbage-collection) whenever new data is added, due to a legacy method of overwriting old hashes with new hashes. However, all of Arbol's newer datasets use an architecture where old hashes are preserved and new posts reference the previous post. This methodology creates a linked list of hashes, with each hash containing a reference to the previous hash. As the length of the list becomes computationally burdensome, the system consolidates intermediate nodes and adds a new route to the head, creating a [DAG (directed acyclic graph)](../concepts/merkle-dag.md) structure. Heads are always stored in a master [heads.json reference file](https://gateway.arbolmarket.com/climate/hashes/heads.json) located on Arbol's command server.
+9. **Garbage collection:** Some older Arbol datasets require [garbage collection](../concepts/glossary.md#garbage-collection) whenever new data is added, due to a legacy method of overwriting old hashes with new hashes. However, all of Arbol's newer datasets use an architecture where old hashes are preserved and new posts reference the previous post. This methodology creates a linked list of hashes, with each hash containing a reference to the previous hash. As the length of the list becomes computationally burdensome, the system consolidates intermediate nodes and adds a new route to the head, creating a [DAG (directed acyclic graph)](../concepts/merkle-dag.md) structure. Heads are always stored in a master [heads.json reference file](https://web.archive.org/web/20230318223234/https://gateway.arbolmarket.com/climate/hashes/heads.json) located on Arbol's command server.
 
 ### The tooling
 

docs/concepts/cod.md

@@ -25,7 +25,7 @@ Bacalhau can:
 - Run against data [mounted anywhere](https://docs.bacalhau.org/#how-it-works) on your machine.
 - Integrate with services running on nodes to run jobs, such as [DuckDB](https://docs.bacalhau.org/examples/data-engineering/DuckDB/).
 - Operate at scale over parallel jobs and batch process petabytes of data.
-- Auto-generate art using a [Stable Diffusion AI model](https://www.waterlily.ai/) trained on the chosen artist’s original works.
+- Auto-generate art using a [Stable Diffusion AI model](https://web.archive.org/web/20250313163631/https://www.waterlily.ai/) trained on the chosen artist’s original works.
 
 ### More Bacalhau resources