Under the hood, Dot Cloud leverages the speed of Python’s asynchronous capabilities to handle concurrent crawling without compromising performance. The project is fully containerized with Docker, making it incredibly easy to deploy and run on any machine with a single command. With a clean architecture that separates the heavy lifting of the crawler from the interactive visualization, Dot Cloud serves as both a practical utility for developers and a fascinating window into the complex web of digital content.

Try it out at https://www.webcloud.benny-water.com

Across industries, large tech and corporate companies are investing heavily in AI-powered chatbots to streamline customer support. But despite the enthusiasm, the reality for many users is far from helpful. Most chatbots today are implemented in ways that do more harm than good.

First, these bots are often poorly trained and lack real problem-solving ability. They tend to function like static FAQ pages, rigidly sticking to pre-set scripts. As noted in AskStylo’s 2024 report: “Chatbots… trap customers in endless loops, forcing them to fight the system just to reach a human” [Stylo, 2024]. This loop often leaves customers more frustrated than when they started.

Second, there’s a fundamental failure to educate users on what the chatbot actually can and cannot do. Instead of a clear introduction, users are often dropped into AI interactions without any context. According to SpurkNow’s 2025 chatbot guide, “setting expectations in the first message” is key to building trust and avoiding confusion [SpurkNow, 2025]. Without this, customers either over-rely on the bot or distrust it altogether.

Worse still, these bots are frequently used as barriers to live support. A 2025 study by CX Today found that 46 percent of users believe chatbots are deliberately used to block access to human agents [CX Today, 2025]. The result? Users churn. When customers feel stonewalled by AI, they are not just annoyed, they are more likely to leave. In the words of CMSWire’s October 2025 analysis, this approach is “a billion-dollar mistake when deployed wrong” [CMSWire, 2025].

Finally, many companies make the strategic error of trying to replace human support too early. Even advanced bots still lack empathy, contextual understanding, and the ability to manage complex queries. Seventy-five percent of consumers agree that current chatbots cannot handle nuanced support issues [Stylo, 2024]. When there is no fast human fallback, this leaves customers stranded.

A Better Way to Deploy AI Support: Clear Roles, Smart Handoffs, Real Choices

The solution is not to abandon AI chatbots, it is to use them more wisely. AI works best not as a replacement, but as a support-layer that augments and accelerates service. The key lies in setting clear boundaries, training bots for the right tasks, and making sure customers always have an easy off-ramp.

Take the example of a fictional telecom provider, let’s call them “UniTel.”

UniTel deploys a support chatbot on its website and app. Crucially, the bot begins every interaction with a clear statement:

“Hi, I’m UniBot! I can help you with common tasks like billing questions, modem resets, and checking for outages. If your issue is more complex or you’re not sure, I can take down your details and get you to the right expert. You can also request to speak to a human at any time.”

Here, the chatbot is doing two critical things:

1. Setting expectations (“here’s what I can help with”)
2. Offering choice (“you can opt out at any time”)

From there, the bot attempts resolution if the issue is simple, say, guiding a customer through resetting their Wi-Fi router. But if the customer says it’s not working, or if the problem lies outside the bot’s scope (e.g., intermittent connection issues across devices), the AI shifts gears.

At this point, the bot clearly informs the user: “This issue appears to be outside of what I can solve directly. I will now ask a few questions to help route your request to the right technical expert.” This transparency reassures the customer that they are being taken seriously and are on their way to real assistance.

Instead of guessing further, the bot switches into triage mode: it asks smart follow-up questions to understand the problem (“When did the issue start? Is it affecting all devices?”) and uses that input to generate a detailed support ticket. That ticket is routed to the right technical team with full context, so the human agent who picks it up does not need to start from scratch. The customer avoids being bounced from queue to queue.

Even better, for those who dislike bots altogether, UniTel retains a small but effective first-line human support team. These agents handle the small share of users who request direct contact upfront. This hybrid model still reduces overhead, but without alienating users.

When deployed this way, AI-assisted support actually enhances customer experience. It speeds up resolution for routine tasks, streamlines routing for complex issues, and gives every customer a sense of control. As Assembled’s 2025 report concluded: “The best AI agents don’t pretend to be perfect. They know when to help, when to ask, and when to get out of the way” [Assembled, 2025].

In the end, successful chatbot integration is not about replacing humans, it is about elevating the support experience. With better training, clearer communication, and frictionless opt-outs, AI support can become a true asset, not a liability.

Sources

• AskStylo, “The Dark Side of Chatbots,” 2024
• SpurkNow, “AI Chatbot Guide,” 2025
• CX Today / NewVoiceMedia, Consumer Survey Report, 2025
• CMSWire, “AI in Customer Service: Billion-Dollar Mistake,” October 2025
• Assembled, “Why Support Teams Are Ditching Chatbots for AI Agents,” March 2025

Core apps I actually use

BetterNews is my custom news and RSS setup. It is built around the idea that reading the news should be fast, searchable, and not driven by engagement algorithms. If you want the deeper breakdown, I wrote up what it is and why I built it here: https://benny-water.com/better-news-rss-news-aggregator-app/

Immich handles photos. It is my “own cloud photos” solution, with machine learning running alongside it for indexing and search. It is one of those services that quietly becomes essential once it is stable.

Invoice Ninja is for invoicing and keeping the money side of projects organized. It is not exciting, but it removes friction, which is the whole point.

EspoCRM is in the lab as a CRM sandbox. It is where I can test workflows and structure before I commit to anything real. I like having a space where experiments do not contaminate production.

Jellyfin is the media server. It is simple, reliable, and fits the rest of the self hosted mindset.

n8n is where automation lives. If something needs to move data between systems, send alerts, or trigger a routine job, it usually ends up here.

Open WebUI, Kotaemon, and Ollama cover the local AI corner of the lab. Open WebUI is the interface, Ollama runs models on device, and Kotaemon is the “bring your own docs” knowledge base style layer. The goal is practical help without sending private data to external services. Some of it is pure productivity, some of it is just curiosity, and all of it stays under my control.

Terminal RPG is a self hosted game project I tinker with. Not everything in a home lab has to be serious. Sometimes the point is simply building something fun and learning along the way.

And then there are the multiplayer staples: Minecraft Server is the reliable classic, while the Hytale Server is there because I like being ready for new worlds when they show up. Both are also a good test for networking, backups, and performance under real use.

Infrastructure that makes the rest possible

On the plumbing side, Traefik handles routing and makes it easier to expose services cleanly without turning my network into spaghetti.

AdGuard Home is the quiet MVP. Network level ad blocking and DNS control makes everything feel calmer and faster.

Dockhand is part of keeping containers manageable without living in the command line 24/7.

PHP Apache is my general purpose web runtime. It is there because sometimes you just need a simple web service without reinventing anything.

Data and the supporting cast

A lot of the lab is only possible because the data layer is solid. I keep the usual trio around: MariaDB, Postgres, and Redis. Between those three, most apps have what they need, and I can choose the right tool without forcing everything into one database shape.

For the AI stack, I also have Pipelines for Open WebUI, which helps structure the “do this, then that” flow when I want repeatable, controllable behavior instead of one off prompts.

Why this setup works for me right now

The theme this month is control and repeatability. Most of these services exist because I want tools that feel stable, understandable, and mine. If something breaks, I can inspect it. If something needs to change, I can evolve it. And if a tool handles sensitive data, I can keep it local by default.

That is January. A mix of practical services, a small AI lab, and a couple of game servers to keep it human.

If you want, I can also turn this into a “Now running” section with a one line description per service, plus a quick diagram style outline you can drop into the post (Home, Core Apps, Infra, Data, AI).

Why Work from Home Is a Net Positive

At this point, the productivity argument is largely settled. Across industries and countries, remote and hybrid workers consistently report equal or higher productivity, better focus, and improved work–life balance. Large-scale research has found that employees who work from home two days a week “are just as productive … as their fully office-based peers.” Moreover, 76% of employees say flexible work arrangements significantly improve their work–life balance. Reduced commuting alone frees up hours of mental and physical energy every week. That energy either goes back into work or into personal recovery, both of which benefit the employer in the long run.

Retention is another overlooked factor. Employees with flexible work arrangements are less likely to leave not because they are more loyal by default, but because flexibility lowers friction in everyday life. When work fits better around health, family, and personal obligations, people stay longer. For example, in a recent randomized experiment at a large company, resignations fell by 33% among workers who shifted from working full-time in the office to a hybrid schedule. Women, non-managers, and those with long commutes were the least likely to quit when office days were reduced. From an employer perspective, this translates into lower hiring costs, more stable teams, and access to a wider talent pool. A U.S. Bureau of Labor Statistics analysis notes that remote work led to lower turnover as job satisfaction rose, substantially reducing firms’ hiring costs. Especially in a European context where cross-border hiring and distributed teams are common rigid office requirements often limit rather than strengthen organizations. Indeed, Stanford’s Nicholas Bloom estimates about 100 million workers worldwide now follow hybrid schedules, enabling companies to recruit talent far beyond the radius of any single office.

In short, remote work works. Not because people work less, but because they work differently.

Where Fully Remote Starts to Break Down

That said, there are real downsides when teams never meet in person. Office culture does not disappear in remote environments, but it becomes thinner. Relationships take longer to form. New employees onboard more slowly. Informal learning and mentoring are harder to replicate. Creative collaboration can suffer when everything is reduced to scheduled calls and shared documents. For instance, a large-scale study at Microsoft found that when everyone shifted to remote work, employees’ collaboration networks became “more static and siloed, with fewer bridges between disparate parts.” This was accompanied by a decrease in synchronous communication and an increase in asynchronous communication, changes that “may make it harder for employees to acquire and share new information across the network.” In parallel, controlled experiments have shown that videoconferencing “inhibits the production of creative ideas” compared to in-person brainstorming. This is not about nostalgia for office life it’s about human dynamics. Trust and team cohesion are still built fastest through shared experiences. Those experiences do not need to happen daily, but they do need to happen at some point.

The problem, then, is not asking people to come into the office occasionally. The problem is asking them to do so without purpose.

The Office Should Offer What Home Cannot

If an employer requires occasional office presence, that presence must feel justified. Too often, employees are asked to commute simply to sit behind the same laptop, attend the same video calls, and perform the same individual tasks they could have done at home. This is where resentment starts. An effective hybrid policy treats office days as a different mode of work, not a continuation of remote work in a different location.

Office days should prioritize:

• Team building and relationship strengthening
• Strategic discussions and workshops
• Brainstorming and creative problem solving
• Onboarding, mentoring, and knowledge sharing
• Shared moments that reinforce company identity

As one workplace expert notes, “regular office time should be used for interactive work: brainstorming, training, team building, and networking.” If those elements are absent, the office becomes a cost rather than a benefit. This caution is echoed by the same study without intentional planning, workers will end up “commuting to just sit at their desks on Zoom calls all day and that will erode morale.”

Making Office Days a Net Positive Experience

This is where many employers miss an opportunity. If you ask people to give up the convenience of working from home, you should give something back. That does not need to be extravagant, but it does need to be intentional. Simple things matter:

• Providing lunch, good coffee, and snacks
• Structuring the day around collaborative sessions rather than solo work
• Creating space for informal conversation
• Ending the day with a team drink or social moment

These gestures are not about bribing employees. They are about signaling respect for their time and effort. They also create shared experiences, which are the foundation of strong teams. From a communication perspective, this framing is critical. When office presence is positioned as a burden or a rule, it fails. When it is positioned as an investment in the team – and backed up by a positive in-office experience – it is often welcomed.

Flexibility and Structure Are Not Opposites

A good work from home policy does not mean the absence of structure. It means being deliberate about where structure adds value. Remote work is highly effective for focused, individual tasks. Offices are effective for connection, alignment, and collective momentum. Research bears this out: a recent field experiment found that a moderate hybrid model about one or two in person days per week delivers an optimal balance between employee well-being and performance. Employees with this schedule reported significantly higher job satisfaction, improved work–life balance, and reduced feelings of isolation compared to peers who were either mostly remote or mostly in-office. Crucially, their performance did not decline; in fact, managers observed modest gains in creativity and quality of work under intermediate hybrid arrangements. By contrast, teams at the extremes (fully remote or five days in-office) had lower satisfaction and no performance advantage over the hybrid group.

Organizations that understand this distinction tend to outperform those that apply a one-size-fits-all policy. The future of work is not fully remote or fully office-based – it is intentional. One Harvard study explicitly warns against rigid mandates, noting that “well-calibrated hybrid systems, not one size fits all mandates, are better suited to fostering employee engagement and organizational resilience.” In practice, this means employers must design work environments both physical and digital around how people actually function, not around how policies used to be written. Done well, hybrid work is not a compromise. It is an upgrade.

Sources:

1. Bloom, N. et al. Hybrid working from home improves retention without damaging performance. Nature, 630, 920–925 (2024). Summary: A randomized trial showed two days WFH weekly improved job satisfaction and cut quit rates by one-third, with no loss in performance.
2. Duarte, R. et al. Finding the Optimal Balance Between Remote and In-Office Work. Harvard CID Research, Nov 6, 2025. Summary: A field experiment found ~2 office days per week maximized satisfaction and maintained performance, whereas fully remote or in-office extremes lowered satisfaction.
3. Yang, L. et al. The effects of remote work on collaboration among information workers. Nature Human Behaviour, 6, 43–54 (2022). Finding: Firm-wide remote work made collaboration networks more siloed, reducing cross-team information sharing.
4. Brucks, M.S. & Levav, J. Virtual communication curbs creative idea generation. Nature, 605, 108–112 (2022). Finding: Videoconferencing hampered creative brainstorming, as remote teams produced fewer creative ideas than in-person teams.
5. Gallup. Hybrid Work Global Indicator (2022). Data point: 76% of employees say flexible arrangements improve their work-life balance.
6. Future Forum (Slack/BCG). Pulse Report – Fall 2022. Insight: Without intentional planning of in-office time for collaboration, employees may end up commuting only to sit on Zoom, hurting morale.
7. Stanford University News. Study finds hybrid work benefits companies and employees (June 12, 2024). Summary: Hybrid schedules had zero impact on productivity or promotions and dramatically improved retention (33% drop in quit rates).

My personal media consumption needed more than a generic aggregator. It needed a coherent ecosystem that filtered noise, meant strictly for local deployment, and transformed information overload into actionable insights.

I architected a full-stack local environment end to end. This included a resource-efficient AI analysis pipeline, a minimalist mobile-first UI, and a granular quality control system. The result is a unified platform that works seamlessly for daily reading, designed to support mindful consumption while running entirely on local hardware.

Content Engine

Hardware-Aware AI Pipeline

To make running local LLMs viable on standard hardware, I engineered a “Fail-Fast” decision pipeline that minimizes GPU overhead. Instead of processing every article fully, the system strictly enforces a tiered analysis order:

Language Check: Immediate discard of non-target languages before scraping.

Scoring Phase (Fast): The AI first reads the raw content to assign a Quality Score and Clickbait Probability. This is a rapid, low-cost operation.

Filtering Gate: If an article falls below the defined Quality Threshold or exceeds the Clickbait Threshold, processing stops immediately.

Rewrite Phase (Intensive): Only the high-quality survivors are passed to the expensive rewriting model to generate summaries and fix headlines.

This approach reduced GPU runtime by approximately 60%, allowing the system to run in the background without impacting daily workstation performance.

Intelligent Transformation

Headline Rewrite & Summary

Beyond simple filtering, the system actively intervenes to reduce cognitive load. I implemented an **AI Headline Rewrite** module that treats the original title as unreliable user input.

Purpose: To de-sensationalize news and neutralize emotional triggers (e.g., changing “You Won’t Believe What X Did” to “X Announces New Policy”).

Implementation: The LLM analyzes the full article body to extract the core factual event, then generates a purely descriptive title.

Quality Control

Granular Filtering Configuration

The system moves beyond simple keyword blocking, relying instead on semantic understanding to curate the feed. I integrated specific, user-tunable controls to define “quality”:

Minimum Quality Score (0-100): A distinct threshold that filters content based on depth, neutrality, and factual density.

Maximum Clickbait Tolerance (0-100%): A probabilistic filter that hides sensationalist or misleading headlines.

Processing Concurrency: Adjustable thread control (1-5 concurrent articles) to balance speed against system heat/fan noise.

Max Article Age: Automated pruning of stale content to ensure relevance.

Diff-Based Change Detection: Custom logic for static sites that only triggers a new feed item when significant HTML changes occur, ignoring footer/layout updates.

Resilience & Failsafes

Recognizing that local AI models can be non-deterministic, I put the user in ultimate control of the data. The system includes manual overrides to recover from model failures:

Force Reprocessing: A dedicated control to invalidate existing analysis and force the AI to re-read and re-score any feed, correcting potential “bad rolls” or hallucinations.

Feed Purge: A destructive failsafe to instantly wipe all cached articles and metadata for a specific source. This ensures that if a feed becomes corrupted or the AI repeatedly misinterprets it, the database can be reset to a clean state without affecting the rest of the system.

Digital Platform

Cleanliness & Mobile-First Design

The interface was stripped down to the essentials to promote focus. I rebuilt the frontend with a strict adherence to usability standards:

Mobile-First Architecture: Large touch targets and bottom-weighted navigation for one-handed use.

Quickread Panel: A dedicated slide-over interface that presents the AI-generated high-density summary. This allows users to consume the “Core Facts” of a story instantly without navigating away to the ad-heavy source URL.

Distraction-Free Reading: A clean, uniform card layout that normalizes content from disparate sources.

Performance: Sub-second load times and efficient DOM management to ensure the app feels like a native tool rather than a web page.

The Czech Pages is more than a directory. It required designing a clear, scalable data model, building a coherent content system and developing a retrieval layer that interprets user intent rather than simply matching keywords. My role spanned UX design, information architecture, front end development and the creation of a semantic search system that helps users find relevant businesses quickly, even with imperfect or ambiguous queries.

This project demonstrates how structured data, clean UX and search logic come together to deliver a genuinely useful local discovery tool.

Foundation

Data architecture and content mode

The core challenge was to create a platform that could grow without becoming chaotic.
I designed a structured data model that defines each listing through consistent taxonomies, metadata and attributes.

Key work included:

• Category taxonomy design
• Standardised metadata fields
• Multi tag classification
• Geo and contextual attributes
• Scalable schema for future categories
• Clean URL and slug conventions for SEO

This foundation enables more advanced filtering, semantic search and future features.

UX Structure

Designing for clarity, speed and minimal cognitive load

A directory succeeds only if users can find what they need with no friction. I built a UX framework prioritising clarity, scanability and predictable interaction.

Key UX decisions included:

• A mobile first layout
• A clear split between browse and direct search
• Card based listings optimised for quick scanning
• Predictable filter structure
• Breadcrumb and navigation logic that avoids dead ends
• Fast access to essential details

The result is a browsing experience that feels light, trustworthy and efficient.

Community Features

A user registration system for a community maintained directory

Because The Czech Pages is designed as a directory maintained by expats for expats, I built a complete front end registration and contribution system. Users can create an account, add listings and manage their own entries through a streamlined dashboard.

Key components:

• Front end user registration
• Secure login and profile management
• User dashboard for creating and editing listings
• Media upload and field validation
• Moderation workflow for all submissions and edits
• Notification flow for approval or revision requests

Listings only go live after moderation, ensuring accuracy and quality across the platform.

Verified owner badge

Business owners needed a way to mark a listing as officially managed without creating a pay to win ranking system. I implemented a verification badge that:

• Confirms ownership
• Helps users trust the information
• Has no impact on search position
• Does not boost visibility
• Respects the integrity of the ranking algorithm, which remains based on category, search intent and data relevance

This approach keeps the platform community driven, fair and transparent.

Semantic Search Layer

Understanding user intent, not just keywords

Traditional directory search breaks down when users type ambiguous or incomplete terms. I designed a semantic search layer that interprets meaning and retrieves the most relevant listings.

Technical work included:

• Mapping categories into semantic clusters
• Handling partial, imperfect and multilingual queries
• Defining fallback behaviours
• Optimising search performance within WordPress constraints

This dramatically improves the quality of results and reduces user frustration.

Visual and Front End Design

A clean, neutral interface that supports the content

The design is intentionally minimal so the content can take priority. The visual system is built around clean typography, generous spacing and an intuitive card based structure.

Work included:

• A neutral colour palette
• Typography for clarity and hierarchy
• Card layouts and iconography
• Smooth transitions and interaction cues

The interface feels modern, trustworthy and easy to navigate.

The Storytellers PR needed a complete refresh of their brand and website with a short turnaround. Their existing presence no longer reflected the quality of their work, and they needed a platform that could communicate clearly, publish quickly and support the agency’s growing portfolio.

I redesigned their identity, built an editorially driven website and developed a custom WordPress backend with a lightweight theme tailored to their content structure. The result is a cohesive system that looks modern, reads with clarity and allows the team to publish new work with very low friction.

Foundation

Brand identity and visual direction

The first priority was to define a clear identity that matched the agency’s strategic and narrative approach. The visual direction is clean, editorial and typography driven, designed to feel premium without being overly ornamental.

Delivered elements included:

• Logo refinement and responsive variants
• A typographic system built around editorial hierarchy
• A neutral yet confident color palette
• Social and presentation templates
• Image and layout rules for consistent storytelling

This provided a stable visual foundation before developing the website.

Digital Platform

Rapid website rebuild with a custom WordPress theme

The client needed a website that could be delivered quickly but still feel highly polished and structured. I built a new site on WordPress with a custom lightweight theme designed specifically for their content model.

Key improvements included:

• Streamlined information architecture focused on narrative clarity
• Modular layout blocks for case studies and newsroom articles
• A homepage designed as a clean editorial introduction
• A simplified WordPress backend so the team can publish without complexity
• Performance and SEO improvements built directly into the theme
• A content structure that is flexible enough for future expansion

The site now works as a practical, modern publishing tool rather than a static showcase.

Editorial and Messaging

A structured voice that fits the agency’s storytelling focus

I refined the core site messaging to ensure consistency with the updated brand identity and the agency’s editorial style.

Deliverables included:

• About page tone refinement
• Service descriptions rewritten for clarity and trust
• Case study structure and headline hierarchy
• Newsroom article formatting guidelines

The results are clean, credible and easy for the agency to maintain.

Barber Studio Don needed more than a website. They needed a coherent ecosystem that connected every touchpoint of their business: online presence, studio materials, booking flow, customer communication and ongoing brand visibility.

I rebuilt their full digital environment end to end. This included brand identity refresh, responsive website development, integrated booking logic, analytics dashboards, and all print materials such as service menus, posters and promotional cards. The result is a unified brand that works seamlessly for both customers and staff, and a system designed to support daily operations while driving measurable growth.

Foundation

Brand identity, visual language and studio materials

The first step was to bring clarity and consistency to the visual identity. The studio received a cohesive design system applied across:

• Service menus
• Price cards
• Window signage
• Posters and seasonal promotional materials
• Loyalty cards
• Social media templates

This ensured that customers experience the same brand tone whether they walk past the window, browse online or book an appointment.

Digital Platform

Website, booking flow and technical foundation

Barber Studio Don needed a website focused on clarity and speed.
I designed and rebuilt the entire platform with a focus on performance, usability and conversion. Core improvements included:

• A clear and focused content architecture
• A mobile first interface built to bring users directly to booking
• A custom booking flow fully integrated with Noona
• SEO focused metadata, structure and schema
• Performance tuning for faster load times and improved ranking

The new site serves as the central hub of the business, linking studio operations, Google Business visibility, social channels and analytics.

Social Visibility

Content planning and communication strategy

To support ongoing customer acquisition and brand memory, I developed a repeatable communication system:

• Reel and post templates
• Seasonal content planning
• Clear messaging guidelines
• Photo treatment and visual consistency
• Studio announcements, promotions and Google Business updates

This allowed the client to continue publishing in a clear, on brand format without needing constant outside support.