What Holiday Extras Did with Codex

Holiday Extras, a travel company, used OpenAI’s Codex to rebuild its main travel app. According to a case study published by OpenAI on November 12, 2026, the company shipped eight releases in two weeks, each with at least three features. The company says this job would previously have taken more than six months.

Codex is an AI tool that can turn natural language instructions into code. Holiday Extras used it to turn customer expertise into better apps, higher booking conversion, and new possibilities for growth. The company’s CEO Matthew Pack, along with David Lee and Liam Thompson, gave colleagues the confidence to build with Codex.

Why This Matters for Small Business Owners

For small business owners, this story shows that AI can help you build tools that understand your customers better and drive growth—no coding team required. Holiday Extras is a 900-person company, but the principle applies to any size business: you can use Codex to turn your own customer expertise into apps that boost conversion.

The key takeaway is speed. Instead of waiting months for a developer to build a feature, you can use Codex to prototype and ship in weeks. This means you can respond faster to customer feedback and iterate on your product more quickly.

How Codex Works in Practice

Codex works by taking short, sharp spoken or written instructions and turning them into working code. At Holiday Extras, employees used voice commands into Codex, Claude, and other tools to build more of their own software. This means you don’t need to be a programmer to create useful apps.

The company’s experience shows that Codex can help non-technical staff build tools that previously required a dedicated development team. For a small business, this could mean building a custom booking system, a customer feedback tool, or a simple inventory tracker without hiring a developer.

What Small Businesses Can Learn from Holiday Extras

First, start small. Holiday Extras didn’t rebuild everything at once—they shipped eight releases in two weeks, each with at least three features. This iterative approach lets you test ideas quickly and adjust based on what works.

Second, focus on customer expertise. The company used Codex to turn what they know about their customers into better apps. For a small business, this could mean building a tool that solves a specific pain point for your customers, like a booking reminder or a personalized recommendation engine.

Third, give your team the confidence to experiment. Holiday Extras leaders encouraged employees to try building with Codex. You can do the same by letting your staff explore AI tools in a low-risk environment.

The Bottom Line for Your Business

Holiday Extras shows that AI can dramatically speed up software development. For a small business, this means you can build tools that were previously out of reach due to cost or complexity. You don’t need a big team or a huge budget to start.

The company’s results—higher booking conversion and faster response to customer feedback—are goals any business can pursue. By using Codex, you can turn your own expertise into apps that drive growth.

Takeaways

• Holiday Extras rebuilt its main travel app in two weeks using Codex, a task that would have taken six months without AI.
• Small businesses can use Codex to build custom tools without a coding team, turning customer expertise into apps that boost conversion.
• Start small, iterate fast, and give your team the confidence to experiment with AI tools.

What Is Waku Agent?

Waku Agent is a local-first personal assistant built around four pillars: Harness, Loop, Memory, and Eval/LLM-Ops. According to the GitHub repository, it is a codebase you own, not a product you use. The loop is about 95 lines of plain Python, and the memory is stored in a single SQLite file at ~/.waku/state.db.

The agent runs entirely on your laptop, with no cloud dependencies. It supports multiple models including Anthropic, OpenAI, Gemini, DeepSeek, MiniMax, Kimi, GLM, OpenRouter, OpenCode Zen, and OpenCode Go. You set the provider and paste one key to get started.

Why Small Businesses Should Care

Small businesses often juggle customer preferences, scheduling, and repetitive tasks. Waku Agent can replace manual note-taking and spreadsheets with a single, readable codebase. For example, you can tell it to remember that a client prefers morning meetings, and it will recall that information even after a restart.

Because the memory is local and stored in a SQLite file, you retain full control over your data. There is no third-party cloud service holding your customer information. This is especially valuable for businesses that handle sensitive data or want to avoid vendor lock-in.

How It Works

The agent’s loop is about 95 lines of plain Python. It reasons, acts, and repeats, with two ways to stop. The memory system includes semantic, episodic, and procedural (skills) types. A gate decides whether to remember, and consolidation decides what to keep.

The dashboard lights up every message as it flows through the harness. You can watch the agent think in real time. The eval system includes deterministic tests and LLM-as-judge side by side, with a release gate. This means you can test changes before deploying them to production.

Getting Started

To install, you run ‘pip install waku-agent’ and then ‘waku’ to talk to the agent in the terminal. You can also run ‘waku dashboard’ to open a browser cockpit at localhost:7777. The first run will tell you which API key to set.

If you want to read or modify the code, you clone the repository and install it with ‘uv pip install -e .’. The agent supports connecting to Waku Memory, a hosted memory service you can share across agents like Claude Code and Codex. You can also connect to live data sources like treg for research.

What Makes It Different

Unlike ChatGPT or Claude Desktop, which are products you use, Waku Agent is a codebase you own. You can read and change the loop, memory schema, gate, and eval harness. Compared to larger open-source assistants like OpenClaw or Hermes, it has the same architecture but with 1/100th the code.

The repository has 2.2k stars and 428 forks on GitHub. It is licensed under MIT for the main code, with the hosted/ folder under Elastic License 2.0. The community is active, with a Discord server and good first issues for new contributors.

Takeaways

• Waku Agent is a local-first AI assistant you own, with a 95-line loop and SQLite memory.
• Small businesses can use it to remember customer preferences, automate tasks, and keep data private.
• It supports multiple models and has built-in eval for testing changes before deployment.

Why Small Businesses Need Local Dictation

Small businesses often juggle multiple tasks and limited staff. Every minute spent typing can be better spent on client interaction or strategic planning. Whisperstream lets you speak instead of type, turning thoughts into clean text instantly. Because the app runs locally, no audio or transcription leaves your office, preserving client confidentiality and avoiding the recurring costs of cloud‑based dictation subscriptions. The built‑in Stream‑1 Mini model automatically removes filler words like “uh” and “um” and applies spoken corrections, so you get polished emails, notes, or reports without manual editing. This speeds up communication and reduces the chance of errors that could affect customer relationships.

Key Features Tailored for Business Workflows

Whisperstream supports 39 languages, covering most global markets. The app detects which program you’re using – Gmail, Slack, Notion, etc. – and formats output accordingly, so a quick Slack message or a formal email comes out with the right tone. The custom dictionary lets you add proper names, product codes, and industry jargon once, and the cleanup engine matches them by sound, ensuring consistent spelling across all documents. For coding teams, Agent Speak provides real‑time voice feedback for Claude Code and Codex, letting developers dictate prompts without leaving the terminal. All features are free and open source, so you can inspect the code for compliance with internal security policies. The optional cloud cleanup can be turned off, and Lockdown Mode adds password protection for saved history, giving extra control over sensitive business data.

How Whisperstream Integrates Into Daily Operations

Setup is straightforward: download the Authenticode‑signed installer for Windows 10/11 or the Mac version for Apple silicon running macOS 14 or later. The default model needs at least 4 GB of RAM; the customizable model recommends 8 GB. After the initial download, the app works offline, only checking for updates once a day without transmitting audio or text. You can install it on every workstation without license limits, and the transcript history is encrypted locally. When you need to dictate a client email, press your custom hotkey, speak naturally, and the Stream‑Mini model cleans up the output in real time. If you change your mind mid‑sentence, simply say the correction and the app applies it instantly. The result is a ready‑to‑send message that reflects exactly what you intended, saving time and reducing the back‑and‑forth of editing.

Cost Savings and Privacy Benefits

Traditional dictation services such as Dragon or cloud‑based solutions charge subscription fees that add up over three years – often exceeding $400. Whisperstream is completely free with no account required, so the total cost is zero. Because the speech‑to‑text processing happens on the device, there is no path for audio to travel to external servers, eliminating the risk of data leaks or unauthorized use. This local processing also means you’re not dependent on internet connectivity for everyday dictation, only for the initial model download and optional update checks. The open‑source nature of the project lets you verify that no telemetry or hidden data collection exists, giving peace of mind for businesses handling sensitive client information.

Getting Started Without Disruption

To begin using Whisperstream, visit the GitHub repository or the official download page and install the Windows version. The installer is Authenticode‑signed, confirming its legitimacy. After installation, configure your hotkey, add any business‑specific terms to the dictionary, and test dictation in a text editor or email client. The app includes a transcript history feature for reviewing past dictations, which is stored encrypted on the local machine. If you encounter issues, you can open a public GitHub issue with logs attached or contact [email protected] for private assistance. The documentation and changelog are publicly available, so you can stay informed about updates without relying on vendor‑provided support tickets.

Takeaways

• Whisperstream provides a free, fully local dictation solution that automatically cleans filler and applies spoken corrections, saving small businesses time and money while keeping all voice data on‑premises.
• The app supports 39 languages, adapts output style per application, and includes a custom dictionary for business terminology, making it versatile for email, chat, notes, and coding workflows.
• Open‑source and privacy‑first design eliminates subscription fees and data‑leak risks, offering a secure alternative to cloud‑based dictation services for small‑business owners.

What Is a Self-Hosted Occupancy Board?

A self-hosted occupancy board is a simple tool that shows everyone in a building and whether they are in, out, or away. Each person chooses what contact information and which picture — photo, company logo, or initials — is shown. The board has five layouts: table, list, cards, tiles, and board. You can switch between them with a layout switcher.

The board is built as one Node.js process plus one SQLite file. It has no dependencies beyond those. It runs on your own server, not on a third-party cloud. The code is public on GitHub, and you can deploy it with a Dockerfile, a compose file, and a systemd unit. The deployment guide is in the repository.

Why Small Businesses Should Care

Small businesses often rely on group texts or hallway conversations to know who is in the office. That can be slow and unreliable. A self-hosted occupancy board gives you a single, glanceable display that updates in real time. You can see at a glance who is in, who is out, and who is away.

The board includes an admin console with a time log, per-person profile pages, and CSV import and export of all people. You can import a list of people from a spreadsheet and export events for record-keeping. The admin console also has a sidebar with overview, tenants, people, and a time log with shifts and hours.

Privacy by Design

The board is built with privacy in mind. Name, title, picture, and status are always shown. Office is shown by default. Phone and email are hidden until the person or an admin turns them on. Search never matches phone numbers or emails. This means anyone who opens the home page can see who is in the office, but personal contact details stay private unless the person chooses to share them.

Each person controls what is shown on their profile. They can sign in at /me to manage their own information. Admins can manage any profile through a dialog. The board also supports tenant logos and a tenant-manager role.

What You Need to Know Before Deploying

The board is self-hosted, so you need a server to run it. The repository includes a Dockerfile and a compose file with an optional Cloudflare Tunnel. The deployment guide explains how to set it up. The live demo runs on GitHub Pages with made-up data stored in the visitor’s own browser.

Two things still need attention: you need to switch on GitHub Pages for the live demo, and you need to decide what to do about real names and phone numbers that are in the public repository. The build log notes that the repository is public, so sensitive data should not be committed.

The board was built by Jordan Mason with Claude Code on November 9-10, 2026. It is merged into main and running on a preview server. The preview is behind Cloudflare Access and uses proxy-aware settings to handle cross-site checks.

How It Compares to Other Options

The build log notes that the team searched similar products — Savance EIOBoard, uSignIn, OnLocation, Creative In/Out — and found no maintained open-source equivalent. Most commercial options are paid and cloud-based. A self-hosted board gives you full control over your data and no recurring subscription fees.

The board is not a Kanban board or a task management tool. It is specifically for showing presence in a physical space. If you need a Kanban board, options like Planka, WeKan, or Kanboard exist. But for a simple occupancy display, this self-hosted board fills a gap.

Takeaways

• A self-hosted occupancy board runs on your own server with one Node.js process and one SQLite file, no external dependencies.
• Each person controls what contact information and picture is shown; phone and email are hidden by default.
• The board includes five layouts, an admin console with time log, CSV import/export, and per-person profiles.

What Is Google Tabletop?

Google has launched an experimental AI tool called Tabletop that generates customized collaborative hubs to help you coordinate with family and friends, according to The Verge. The platform shares visual similarities with workplace management apps like Asana and Monday, but includes a built-in AI feature that lets you describe personalized interactive tools—such as savings trackers, chore charts, or event plans—and the AI builds them for you.

Google Labs design and engineering lead Roman Nurik explained in a blog post: “Just add in your links, files and ideas, then prompt @tabletop — or pick a starter template — to generate custom interactive trip itineraries, dinner polls, expense splitters and tournament brackets.” Once your space is ready, you drop an invite link in your group chat so everyone can use it at the same time, like a Google Doc.

How It Works for Small Businesses

For small business owners, Tabletop offers a way to quickly set up internal coordination dashboards without any coding. Instead of buying or building a project management tool from scratch, you can describe what you need—like a shared expense tracker for a team project or a chore chart for a small retail store—and Tabletop generates it instantly.

The tool is currently in beta and accessible via a waitlist. Access is limited to people over 18 in the US, according to The Verge. While the examples given focus on personal planning, the same functionality applies to business use: creating custom tools for tracking budgets, planning events, or managing tasks among a small team.

Key Features That Matter to Business Owners

Tabletop combines real-time collaboration with an AI agent that builds custom tabs and controls inside a shared space, as reported by AlphaSignal. The basic workflow is simple: create a space, add links, files, notes, or other source material, then prompt @tabletop to generate the interactive tools you need.

The platform allows multiple people to use the space at the same time, similar to a Google Doc. This means a small team can simultaneously update a shared budget, vote on a dinner poll for a client meeting, or track expenses for a group trip—all without switching between different apps.

What This Means for Your Business

For small business owners who don’t have a developer on staff, Tabletop removes the barrier to creating custom collaboration tools. Instead of spending time learning complex software or paying for expensive subscriptions, you can describe your needs in plain language and get a working tool in seconds.

The tool is still experimental, so it may have limitations. But the fact that Google is building this into a publicly accessible platform signals a shift toward AI-powered, no-code solutions for everyday business coordination. If you’re tired of juggling spreadsheets and sticky notes, Tabletop could be worth trying once it’s available.

Takeaways

• Tabletop lets you describe custom tools like expense splitters or chore charts, and the AI builds them without coding.
• It works like a real-time collaborative space, similar to Google Docs, so your team can update tools simultaneously.
• Access is currently limited to US adults via a waitlist, but the beta is rolling out now.

What NitroMTA Is

NitroMTA is an open‑source mail transfer agent that serves as a drop‑in replacement for Amazon Simple Email Service (SES) v2. The project is hosted on GitHub (GitHub) and is released under the Apache‑2.0 license. It is written in Rust and ships as a single binary that stores its data in SQLite, runs as a systemd unit, and includes a command‑line interface for administration.

SES Compatibility and Migration

The engine implements the SES v2 API surface: SendEmail, SendBulkEmail, email templates, identities, configuration sets, tenants, the suppression list, dedicated IPs, and a subset of SES‑shaped events via SNS. Because the API is identical, moving an application from AWS to NitroMTA only requires changing the endpoint URL and the access keys. The official SDKs for Python (boto3) and other languages continue to work without code changes.

Performance and Reliability

Benchmarks run on an AWS c7gd.metal server with four CPUs show NitroMTA delivering 3,313 messages per second, or 1.169 ms per message. That rate is 3.16 times faster than KumoMTA (1,048 msg/s, 2.492 ms) and 4.11 times faster than Postfix with OpenDKIM (806 msg/s, 4.969 ms). Across nine test runs the engine sustained a median of 1.35 million 16 KiB messages, each DKIM‑signed and TLS‑encrypted, with every message delivered once and verified. Durability is built in: a submission is only acknowledged after it is committed to disk, allowing the queue to survive crashes, power loss, or host failure as long as the storage remains intact. The engine also includes automatic warm‑up per receiving domain, back‑off on complaints or throttling, per‑provider connection limits, and suppression of hard bounces and spam complaints. All outbound mail is DKIM‑signed, sent over TLS, and validated with MTA‑STS and DANE where available.

Running NitroMTA on Your Own Infrastructure

Installation is available as native packages for Debian/Ubuntu (.deb), RHEL/Fedora (.rpm), Arch (AUR), Docker images, Homebrew for macOS/Linux, and generic x86_64/arm64 archives. Each package is signed by GitHub, providing provenance. The setup process consists of six steps, ending with the “nitromta doctor” command that checks reverse DNS, SPF, DKIM, DMARC, and the bounce MX record, then tells you what DNS records to publish. A capture mode lets you run the full engine locally and view each message in a browser without needing DNS or public IPs, useful for testing. To migrate, you keep your existing SES v2 code, point the SDK endpoint at the local engine (e.g., https://127.0.0.1:4443), and use the access keys generated by the NitroMTA admin CLI. The engine handles SMTP delivery, so you still need outbound port 25 open and a reverse DNS name for each sending IP.

Cost, Control, and Suitability for Small Businesses

Because NitroMTA runs on your own hardware or cloud VM, you avoid per‑message fees charged by AWS. You also gain full control over the IP addresses used for sending, which can improve reputation and deliverability when managed correctly. The built‑in warm‑up logic removes the need for a separate IP‑warming service. The solution is most appropriate for businesses that already send a moderate to high volume of email and have the ability to manage a Linux host with port 25 access. For low‑volume senders, the operational overhead may outweigh the cost savings, but the open‑source nature means you can try it in capture mode at no cost.

Takeaways

• NitroMTA provides drop‑in SES v2 compatibility, so you can switch endpoints without rewriting code.
• Performance benchmarks show over 3,000 msg/s on modest hardware, with built‑in durability and deliverability features.
• Running the engine yourself eliminates per‑message AWS fees and gives you direct control over sending IPs and warm‑up.

Happy Veterans Day, Rapid City. This morning we pointed our AI agent, Hermes, at one job: find every Veterans Day special, free meal, and discount it could track down for veterans in Rapid City. It came back with 56 of them. Then we did the part that matters — we checked every single claim before publishing it. Below is the list that survived.

How we did it

Hermes compiled this morning’s list from local news coverage, national Veterans Day roundups, and merchant pages — 56 entries in all. Then each one had to pass three checks:

• The claim check. The deal has to actually appear in a published source, and every entry below links to the source we verified it against.
• The local check. The business has to actually operate in or near Rapid City. Several of our sources are national roundups, which name chains that have never opened a store in South Dakota. 21 entries failed here because we couldn’t confirm a Rapid City location — including a well-known California burger chain with no South Dakota location at all.
• The clock check. The offer has to still be redeemable when you read this. Two deals (Denny’s and Texas Roadhouse, whose lunch windows closed at noon and 2 p.m.) had already ended by publication time, and the VFW’s luncheon, which followed the 11 a.m. parade, was most likely over (no end time was published).

Of the 56 entries Hermes found, 31 made the cut. We cut 21 we couldn’t confirm in Rapid City, 2 whose windows had already closed, 1 event that was most likely already over, and 1 national program we couldn’t tie to a specific participating location in Rapid City. What’s left is below.

Free meals & restaurant deals

• Applebee’s — Free dine-in entrée from the Veterans Day menu. Proof of service required. 2160 Haines Ave. Source: VA News
• Golden Corral — Veterans appreciation night today, 4–9 p.m. The national offer is a free buffet and beverage, no proof of service required. Rapid City. Source: Rapid City Journal · Freebie Depot
• IHOP — Free Red, White & Blueberry Pancake Combo, 7 a.m.–7 p.m., dine-in, proof of service. 550 Disk Dr. Source: Food Network
• Chili’s — Free entrée from the Veterans Day menu, dine-in, ID required. 2125 Haines Ave. Source: VA News
• Red Robin — Free Red’s Big Tavern Burger with bottomless fries, dine-in, ID required. Rapid City. Source: VA News
• Red Lobster — Free Shrimp & Chips meal, dine-in, ID required. 120 Disk Dr. Source: VA News
• Buffalo Wild Wings — Free 10 boneless wings and fries, dine-in, proof of service. Rapid City. Source: Freebie Depot
• Outback Steakhouse — Free Bloomin’ Onion and a beverage, running Nov. 10–11. Rapid City. Source: Freebie Depot
• Cracker Barrel — Free Sunrise Pancake Special, dine-in, proof of service. Rapid City. Source: Freebie Depot
• TGI Fridays — Free meal from a select menu, dine-in 11 a.m.–7 p.m., military ID required. 2205 N Lacrosse St. Source: USA Today, via Rapid City Journal (national roundup; Rapid City address from our research log)
• Ruby Tuesday — Free meal from the Everyday Fresh Menu, dine-in. 821 Fairmont Blvd. Source: VA News
• Freddy’s Frozen Custard & Steakburgers — Free combo meal card when you visit today, redeemable through Nov. 25. Rapid City. Source: VA News
• Wendy’s — Free Dave’s Single, no purchase necessary. 1911 N Haines Ave & 701 E North St. Source: Freebie Depot
• Chipotle — Free chips with guacamole, queso, or salsa with a $5 purchase, ID required. Rapid City. Source: VA News
• Starbucks — Free tall (12 oz) hot or iced brewed coffee for veterans, active duty, and military spouses. Rapid City locations. Source: VA News
• Casey’s — Free coffee or fountain drink, plus a free car wash where available. Multiple area locations. Source: Freebie Depot
• Dunkin’ — Free donut, while supplies last, no ID required. Rapid City. Source: Freebie Depot
• Scooter’s Coffee — Free medium drink, military ID required. Rapid City locations. Source: Freebie Depot
• Circle K — Free medium hot or iced coffee, ID required. Rapid City locations. Source: VA News

Retail discounts

• Kohl’s — 30% off in-store purchases. Today, Nov. 11, is the last day; ID required. Source: VA News
• Target — 20% off one qualifying purchase for Target Circle members with verified military status. Ends today, Nov. 11. Source: VA News
• Walgreens — 20% off eligible regular-price items in store, for veterans, active duty, and their immediate families. Source: VA News
• Tractor Supply — 10% off for verified Hometown Heroes members. Source: VA News

Services

• Rapid Wash Express Car Wash — Free Titanium Wash, 7 a.m.–7 p.m., for active-duty and retired military with ID. 1735 Haines Ave & 5110 Fifth St. Source: Freebie Depot
• Meineke Car Care Center — 10% off most services for active duty and veterans with valid ID (their standing military discount). 1224 E Highway 44. Source: Meineke
• Great Clips — Free haircut today, or pick up a free haircut card to use on a later visit. Participating locations. Source: VA News
• Sport Clips — Free or discounted haircut at participating locations (terms vary by shop), valid ID required. Source: VA News

Events & attractions

• American Legion Post 22 — Free turkey dinner for veterans, 4–6 p.m. (Non-veteran adults $10, kids $5.) 818 E St Patrick St. Source: American Legion Post 22 · KOTA
• Performing Arts Center of Rapid City — Free Veterans Day concert at the Historic Theater, 7 p.m. Free tickets required; reserve through the Performing Arts Center (https://app.arts-people.com/index.php?show=366548). 601 Columbus St. Source: The Rapid City Post
• National Park Service — Fee-free day at national parks. Badlands National Park, which normally charges $30 per vehicle, waives its entrance fee today. Wind Cave National Park has no entrance fee (fees apply only to cave tours). Mount Rushmore entry is always free; its parking fee still applies. Source: National Park Service
• Rapid City Rush — military tickets for the weekend games against the Tahoe Knight Monsters at The Monument — Friday, Nov. 13, and Saturday, Nov. 14, which is Veterans Appreciation Night (7:05 p.m.). The Rush’s promo schedule confirms the Nov. 14 date; we couldn’t load the ticket page to confirm the price or whether it applies Friday, so check with the box office. Not tonight: tonight’s home game is listed as Waggin’ Wednesday. 444 N Mt Rushmore Rd. Source: Rapid City Rush

A quick honesty note

Every offer above is for today, Wednesday, Nov. 11, 2026, unless a different date is stated. Terms, hours, and participation are set by each business and can vary by location, so a quick call ahead is worth it — and bring proof of service wherever it’s noted. We verified each claim against its source this afternoon, but businesses have the final say at the register.

To every veteran in the Black Hills: thank you for your service. We hope this list helps you enjoy a day that’s yours.