When Data Fades: What Cambridge’s “Future Nostalgia” Teaches Us About Backups

Cambridge University is in a race against time. This should sound familiar to any business owner who’s ever lost a file.

This month, the University Library launched “Future Nostalgia.” This is a year-long initiative. It is dedicated to rescuing decades of data trapped on obsolete floppy disks before it’s lost forever. The project helps people recover everything from forgotten WordPerfect files to fragments of Stephen Hawking’s early research.

It may sound like a nostalgic tech rescue mission. However, it highlights a far more urgent truth: digital data doesn’t last forever.

When Data fades, Backups keep history alive.

When Data Storage Becomes Data Loss

Magnetic coatings decay. Drives disappear. File formats go extinct.
Cambridge’s archivists face the same challenges your business will—just on older hardware.

Their solution? Specialized recovery tools like KryoFlux and Greaseweazle. These tools sample the raw magnetic signal itself. This process rebuilds files that would otherwise be lost. It’s a process so delicate that even minor damage can mean the difference between recovery and permanent loss.

For small and mid-sized businesses, that’s a sobering reminder. If your data only exists in one place, it’s already at risk. The risk increases if it’s stored on aging drives.

History Repeats Itself: Unless You Back It Up

Cambridge is racing to preserve digital history. In the same way, modern companies must preserve their operational history. This includes emails, invoices, customer data, and contracts.

Here’s what every business can learn from Cambridge’s mission:
Technology Ages Fast: What seems “safe” today might be unreadable tomorrow.
☁Backups Need Testing: A backup that hasn’t been tested might as well not exist.
Formats Matter: Proprietary files can lock away your data. Old apps may also limit access. Aging servers can make data retrieval as challenging as using a floppy disk.

That’s why SofTouch Systems enforces a three-layer backup policy for every client:

  1. Local Backups – For fast recovery from minor issues.
  2. Cloud Copies – Offsite, secure, and ransomware-resistant.
  3. Verification & Testing – Monthly recovery checks ensure your data actually restores.

Because when data vanishes, recovery becomes a race against physics, not just time.

The Future of Nostalgia Is Prevention

Cambridge’s archivists are doing incredible work preserving the past.
But your business’s focus must be on protecting the present—and the future.

Every invoice, design, or client file you lose today could be your “digital archaeology” project tomorrow. Don’t wait for your own floppy-disk moment.

Start with a free Backup & Recovery Checkup at SofTouchSystems.com
We’ll make sure your data stays recoverable—not just remembered.

How to Make Your First AI Chatbot (Using Free Tools Like Hugging Face and Google)

Artificial intelligence isn’t just for big tech companies anymore. Small businesses, nonprofits, and digital nomads now have the ability to build their own AI chatbots for free. They can use platforms like Hugging Face, Google Colab, and Google Sheets for this task. These tools let you create smart, responsive assistants. They can automate customer support, gather leads, or even teach your clients. All of this is possible without writing a single line of code.

If you’ve ever wanted your own branded AI assistant to answer questions on your website, now’s your chance.

Step 1: Define Your Chatbot’s Purpose

Before touching any tools, define what your chatbot should do.
Ask yourself:

  • What common questions do clients ask again and again?
  • Should the bot collect contact info, provide IT advice, or walk users through troubleshooting?
  • Do you want it to sound friendly, professional, or like your brand’s tone?

Pro tip from STS: Keep the goal narrow for your first version. A focused chatbot is easier to train, test, and improve.

Step 2: Create a Free Hugging Face Account

Go to HuggingFace.co and create a free account. Hugging Face is a community of AI models and spaces that let anyone host apps and chatbots in the browser.
Once signed in:

  1. Click your profile → “New Space.”
  2. Choose Gradio as the app type (perfect for no-code interfaces).
  3. Name your project (e.g., “MyFirstChatbot”) and set visibility to Public.

You now have your own mini app environment ready to host your chatbot—completely free.

Step 3: Add an AI Model

Next, you’ll choose a model to power your chatbot’s “brain.”
Free options include:

  • facebook/blenderbot-400M-distill – general conversation.
  • mistralai/Mistral-7B-Instruct – open-source model great for business Q&A.
  • google/gemma-2b – lightweight and efficient for small Spaces.

Hugging Face will automatically connect the model to your app. You can test right in the browser—no setup required.

Step 4: Build a Simple Chat Interface

Still in your Space, click “Files and Versions,” then “Edit App.”
Paste in a few lines of Python (provided in their docs). Here’s a simplified version:

import gradio as gr
from transformers import pipeline

chatbot = pipeline("text-generation", model="mistralai/Mistral-7B-Instruct")

def respond(message, history):
    reply = chatbot(message, max_length=200, do_sample=True)[0]['generated_text']
    return reply

gr.ChatInterface(respond, title="My First AI Chatbot").launch()

Click “Commit changes.” Your chatbot will go live in minutes—free hosting included.

Step 5: Teach It About Your Business (Using Google)

To make your chatbot useful, you can teach it your business FAQs using free tools like:

  • Google Sheets – store your Q&A data (Question | Answer).
  • Google Colab – a free notebook where you can link your Sheets data and train a small retrieval system.
  • Hugging Face embeddings – use sentence-transformers to make your bot “remember” key topics.

This allows your chatbot to respond to business-specific queries like:

“What services does SofTouch Systems offer in San Antonio?”
“How can I reset my 1Password vault?”

You can gradually improve accuracy by refining your FAQ sheet—no retraining needed.

Step 6: Deploy on Your Website

Once your chatbot works on Hugging Face:

  1. Click “Embed this Space.”
  2. Copy the iframe code.
  3. Paste it into your website or blog’s HTML.

Visitors can now chat directly with your AI assistant 24/7.
For example, SofTouch Systems could embed a chatbot to help clients explore the “No Surprise IT” packages. Clients could also schedule a security audit.

Step 7: Keep It Secure and Ethical

Even free chatbots should follow best practices:

  • Never share private data in public Spaces.
  • Use a custom system prompt reminding users of data limits.
  • Monitor logs for inappropriate queries or security issues.

STS recommends pairing any AI tool with 1Password for secure credential management. Use Bitdefender for endpoint protection too. This is exactly how we do in our Cyber Essentials bundles.

Bonus: Expand Your Chatbot’s Powers

When you’re ready to scale up, you can integrate:

  • Google Drive or Sheets API for data collection
  • Zapier for automated email follow-ups
  • Mailchimp or HubSpot for lead nurturing

These integrations turn your chatbot from a novelty into a full sales assistant.

Conclusion

Building your first AI chatbot no longer requires a development team or big budget. With Hugging Face and Google’s free ecosystem, you can launch a functioning AI assistant in under a day.

At SofTouch Systems, we believe tools like these democratize automation—helping small businesses and digital nomads compete with enterprise-level efficiency.

Ready to try it yourself?
Start building today at huggingface.co. If you’d like help turning your prototype into a branded business assistant, reach out to SofTouch Systems. They offer an AI Integration Services consultation.

Google num=100 Parameter Removal: What It Means for Your SEO in the AI Age

What Just Happened?

If you track search performance, you felt it. The Google num=100 parameter removal in September 2025 changed how results pages load. It also altered how tools collect data. Google confirmed it no longer supports a results-per-page parameter; in practice, standard pages now show about ten results, not 100. As a result, many sites saw impression counts fall and average position “improve” overnight, even though rankings didn’t actually change.

What changed—plain English

Until mid-September, adding &num=100 to a Google results URL forced 100 listings on one page. Rank-tracking tools and scrapers leaned on that trick; so did some power users. Google stopped honoring it. Industry data shows steep drops in Search Console impressions. These drops are noted in tracked keywords across most properties after the switch. Importantly, the change affects measurement, not your real-world visibility.

Why would Google do this? Several reasons line up:

  • Reduce automated scraping and abuse. Killing num=100 raises the cost of bulk SERP harvesting.
  • Nudge pros to official APIs. Expect heavier reliance on Search Console and Custom Search APIs for data access.
  • Align with answer-first interfaces. Fewer paginated links complement Google’s AI-overview direction. (Industry inference from coverage and statements.)
Google Just killed &numb=100: What it means for SEO in 2025

The impact you’ll actually see

From San Antonio to Corpus and the I-35 corridor, our SMB clients reported the same pattern:

  • Impressions down; average position up. Deep results stopped inflating impression totals; averages look “better” without a ranking lift.
  • Fewer “visible” keywords in tools. Trackers can’t grab 100 at a time; coverage looks thinner unless they paginate more (which costs more).
  • Clicks and conversions steady. Because user behavior didn’t swing, real business metrics didn’t nosedive.

What Texas SMBs should do next

  1. Re-baseline your dashboards.
    Reset “normal” levels for impressions, avg position, and keyword counts from September 2025 forward. Add an annotation in your reporting.
  2. Prioritize top-10 outcomes, not page-3 vanity.
    With fewer results per view, page-one coverage matters even more. Refocus content plans on the queries where you can credibly win the top 3–5 spots.
  3. Measure what pays the bills.
    Shift weight from impression totals to CTR, qualified clicks, leads, and revenue. Build monthly reviews around those KPIs, not around keyword inventory swings.
  4. Ask your tool vendors hard questions.
    How are they adapting collection methods? Are they paginating more, sampling differently, or leaning on official APIs? Demand a methodology note.
  5. Tune content for “answer engines.”
    Structure posts with crisp questions, scannable subheads, summaries, and schema (FAQ, HowTo, Product, Organization). That format feeds AI overviews and rich results. (Industry best practice consistent with the shift.)
  6. Fortify entity and brand signals.
    Strengthen E-E-A-T: author bios, local citations, consistent NAP, and first-party references. In an answer-first world, trusted entities surface more often.
  7. Diversify traffic sources.
    Grow email, referral, and social so a single SERP tweak doesn’t derail pipeline. (Common recommendation as data access tightens.)

What this signals about the future of SEO in the AI era

  • From lists to answers. Google continues to compress choices and elevate summaries. Content that directly answers intent—and marks it up—wins more often.
  • Data will be more gated. Scraping shortcuts will fade. APIs, partnerships, and modeled data will power reporting. Expect some tools to raise prices or reduce depth.
  • Brand and local authority rise. When users see fewer links, trust matters more. Texas-relevant case studies, reviews, and citations become durable moats.
  • UX signals tighten the loop. Engagement and satisfaction correlate with rankings when competition intensifies on page one. Keep pages fast, clear, and persuasive.

How STS will protect your pipeline

  • “No-Surprise” reporting. We’ll annotate September 2025 forward, recalibrate baselines, and center reviews on leads and revenue—not noisy impression totals. (Internal alignment with our reporting standards.)
  • Answer-optimized content sprints. We’ll refactor priority pages with FAQs, concise takeaways, and structured data to support rich results and AI picks.
  • Local proof first. Every campaign gains Texas-centric proof assets: case metrics, ticket MTTRs, backup success %, and security outcomes that convert skeptical buyers.

External source links

  • Google confirms results-per-page parameter isn’t supported (Search Engine Land).
  • Industry data on impressions/keyword visibility drops (Search Engine Land, LOCOMOTIVE).
  • Practitioner analyses of the change (Brodie Clark, ZEO).
  • Rank-tracking/collection changes post-num=100 (SerpAPI).

TL;DR for busy owners

The Google num=100 parameter removal didn’t tank your rankings; it changed what tools can see. Re-baseline reports, double down on page-one content, and measure leads and revenue. We’ll handle the technical shifts, and we’ll keep your pipeline predictable in an AI-first search landscape.