More from Founder's blog
Our new Microsoft Teams app just passed Microsoft's certification and is live in the Teams Store: Jitbit Helpdesk Bot. Search "Jitbit" in the Teams app store and it's right there - no custom app uploads, no admin-center gymnastics. Unlike our old webhook integration (one-way channel notifications, now deprecated), this is a proper two-way bot: End users create tickets in chat - type new and get a real form with category and priority, or turn any Teams message into a ticket from the "..." message menu. my tickets lists your open requests Technician replies come back to Teams - as cards with an inline answer box, so the whole conversation can happen without opening the helpdesk Channels get actionable ticket cards - @mention the bot, type subscribe, and every new ticket shows up with Take it / Reply / Close buttons. Cards update in place when the ticket changes - whether that happens in Teams or in the web app Setup takes about two minutes: install the app, click "Generate pairing command" on the helpdesk admin page, paste the command into a chat with the bot. Users are matched by their Microsoft 365 email automatically. Self-hosted customers run their own private bot, so ticket traffic never touches our servers. The bot is included with all plans. Details and setup steps: the integration page and the documentation.
Help desk interviews usually test two things at once: whether you understand the technical basics, and whether you can explain those basics to a person who may already be annoyed, confused, or late for a meeting. That combination matters. A support technician who can diagnose a network issue but makes the user feel stupid will struggle. So will a friendly person who guesses randomly at technical problems. The best candidates show both: methodical troubleshooting and calm, useful communication. Below are common help desk and desktop support interview questions, with sample answers you can adapt to your own experience. Technical Help Desk Interview Questions 1. Can you tell me about yourself? Keep the answer focused on the job. Mention your IT training, support experience, certifications, customer service background, and the kind of technical problems you have handled. Avoid turning it into a life story. A good answer gives the interviewer several useful follow-up paths. For example: "I have been building my support skills through Windows troubleshooting, networking fundamentals, and customer-facing work. I enjoy breaking problems down, documenting what I find, and helping users get back to work without making the process more stressful for them." 2. A user says their monitor is not working. What do you check first? Start with the simple physical checks before assuming a complicated failure. Confirm that the monitor has power, the brightness is not turned all the way down, the video cable is connected securely, and the computer itself is powered on. If the monitor has multiple inputs, make sure the correct input is selected. If those checks do not solve it, continue with another cable, another monitor, or another port. From there, you can investigate graphics drivers, docking stations, sleep state problems, or hardware failure. 3. What is Safe Mode used for? Safe Mode starts Windows with a limited set of drivers and services. It is useful when a normal startup fails, when a bad driver causes crashes, or when you need to remove unwanted software that loads during a standard boot. The exact steps depend on the Windows version, but the general idea is to restart into the recovery or advanced startup options and choose Safe Mode or Safe Mode with Networking. 4. What is an IP address? An IP address identifies a device on a network so other devices know where to send traffic. On a typical office network, a computer may receive its IP address automatically from DHCP, although servers, printers, and network equipment often use fixed addresses. On Windows, you can check the assigned address with ipconfig or ipconfig /all in Command Prompt. You can also find it in the network adapter settings. 5. What is a default gateway? A default gateway is the device a computer uses when it needs to reach something outside its local network. In many offices and homes, that gateway is a router or firewall. Without a working gateway, a computer may still talk to nearby devices but fail to reach the internet or other remote networks. 6. What is Active Directory? Active Directory is Microsoft's directory service for managing users, computers, groups, permissions, and policies in a Windows domain environment. In practical help desk work, you may use it to reset passwords, unlock accounts, check group membership, or confirm which computers belong to the domain. 7. What is a windows domain? A domain is a centrally managed group of users, computers, and resources. Instead of each PC having completely separate local accounts and permissions, a domain lets administrators manage access from one place, usually through Active Directory. For help desk work, this matters because a user's ability to sign in, access file shares, use printers, or launch certain applications may depend on their domain account and group memberships. 8. A printer is printing strange symbols or garbled pages. What might cause that? A common cause is the wrong printer driver or a corrupted print job. I would first clear the print queue, confirm the correct printer model and driver, and reinstall or update the driver if needed. I would also check whether other users are affected, because that helps determine whether the issue is local to one workstation or shared across the printer or print server. 9. What are common Ethernet cable categories? Common twisted-pair Ethernet cable categories include Cat5e, Cat6, and Cat6a. Cat5e is widely used for gigabit networking, Cat6 is also common for gigabit and shorter 10 GbE runs, and Cat6a is designed for 10 GbE over longer distances. In an interview, it is usually enough to show that you understand the categories affect supported speed, distance, and installation quality. 10. What is a blue screen? A blue screen, often called a BSOD, is a Windows stop error. It can be caused by failing hardware, bad drivers, memory problems, disk issues, overheating, or low-level software conflicts. A sensible troubleshooting path is to note the stop code, check recent changes, review Event Viewer, update or roll back drivers, run hardware diagnostics, test memory, and look for patterns such as crashes only after docking, printing, or launching a specific application. 11. What does DHCP do? DHCP stands for Dynamic Host Configuration Protocol. It automatically gives network settings to devices, including IP address, subnet mask, default gateway, and DNS servers. Without DHCP, users or administrators would have to configure those values manually on each machine. 12. What does DNS do? DNS translates names people can read into IP addresses computers can use. For example, when someone visits a website, DNS helps find the server behind that domain name. In support work, DNS problems can look like "the internet is down" even when the network connection itself is working. Useful checks include trying another site, using nslookup, checking the DNS server address, and flushing the local DNS cache when appropriate. 13. What is a VPN? A VPN creates an encrypted connection from a user's device to a private network. Remote employees often use VPNs to access internal applications, file shares, intranet sites, or administrative systems that are not exposed to the public internet. When troubleshooting VPN issues, check credentials, MFA prompts, internet connectivity, client version, certificate problems, and whether the user is on a restricted network. 14. What is the ping command used for? ping sends test packets to another host and reports whether replies come back. It is a quick way to check basic reachability and latency. It does not prove that a website, file share, or application is working, but it can help narrow down whether a device can reach another device at all. 15. What is Group Policy? Group Policy is a Windows feature used to apply settings across users and computers in a domain. Administrators can use it to configure password rules, mapped drives, desktop restrictions, security settings, software deployment, browser settings, and many other policies. For a help desk technician, Group Policy is often relevant when a user cannot access a feature that someone else can, or when settings keep changing back after a reboot or login. 16. What is a PST file? A PST file is an Outlook data file, commonly used to store email, calendar items, contacts, and archives locally. In support scenarios, PST files come up during Outlook migrations, archive recovery, mailbox troubleshooting, and storage cleanup. 17. How would you change folder permissions? First, confirm who should have access and whether the folder is local, on a file server, or controlled by a broader policy. On Windows, folder permissions are usually managed from the Security tab in the folder properties, where an administrator can add users or groups and assign permissions such as Read, Modify, or Full Control. In a company environment, it is usually better to grant access through security groups rather than adding individual users one by one. 18. What is the difference between a hub and a switch? A hub sends traffic out to every connected device, whether the traffic is meant for that device or not. A switch is smarter: it learns which devices are connected to which ports and forwards traffic only where it needs to go. That makes switches much more efficient and secure for modern networks. Hubs are mostly obsolete, but interviewers still ask the question because it tests basic networking knowledge. 19. How would you recover files from a virus-infected computer? Do not rush to copy files from an infected machine onto the network. First isolate the computer, document symptoms, and follow company security procedure. If recovery is approved, use a clean, trusted environment and scan the drive with updated security tools before moving any files. Depending on the incident, the correct answer may involve escalating to security, preserving evidence, or restoring from a known-good backup rather than manually extracting files. 20. Why should we hire you? Use this answer to connect your skills to the role. Mention technical fundamentals, reliability, communication, willingness to learn, and any experience that proves you can handle real users under real pressure. A strong answer is specific: "You should hire me because I can troubleshoot methodically, explain technical issues clearly, and stay patient when users are stressed. I also document my work, ask for help before wasting time, and keep learning so I can solve more problems independently." Customer Service Help Desk Interview Questions Technical answers matter, but help desk work is still service work. Interviewers want to know how you behave when the problem is unclear, the user is frustrated, or the ticket queue is already full. 1. What makes someone good at help desk work? A good help desk employee listens carefully, asks clear questions, and explains the next steps without drowning the user in jargon. They also know when to keep troubleshooting and when to escalate. Speed matters, but accuracy, documentation, and the user's experience matter too. 2. What do you do when you cannot solve an issue? I would gather the important details first: what changed, who is affected, what error appears, what has already been tried, and how urgent the issue is. If I still cannot resolve it, I would escalate with clean notes so the next person does not have to start from zero. 3. How do you handle an angry or frustrated user? Stay calm and avoid arguing. Let the user explain the problem, acknowledge the impact, and move the conversation toward the next useful action. A simple sentence like "I can see why that is frustrating; let's check the fastest things first" can lower the temperature without making unrealistic promises. 4. Tell me about a conflict you had and how you resolved it. Choose an example where you stayed professional and solved the actual issue. The best answers are not dramatic. A misunderstanding about priority, a handoff problem, or a disagreement about procedure can work well if you explain what you learned and how you prevented it from happening again. 5. How would you rate your troubleshooting ability? Be confident but realistic. If you are early in your career, a four out of five is often more believable than claiming perfection. Explain that you are strong at structured troubleshooting, documentation, and asking good questions, while still knowing when to escalate unfamiliar problems. 6. How do you keep your IT knowledge current? Mention specific habits: reading vendor documentation, practicing in a lab, following release notes, taking courses, studying for certifications, or learning from tickets after they are resolved. Interviewers like answers that show steady curiosity rather than vague "I read online" statements. 7. Why do you want to work in help desk? A good answer connects problem-solving with service. You might say that you enjoy figuring out technical issues, helping people get unstuck, and learning a wide range of systems. Help desk roles expose you to many parts of IT, which makes them a strong starting point for a support career. 8. How do you stay organized? Talk about your system. For example, you might prioritize tickets by urgency and impact, keep notes inside the ticket, use reminders for follow-ups, and close the loop with users before marking work complete. The point is to show that you do not rely on memory alone. 9. Is teamwork important in help desk? Yes. Support teams depend on clean handoffs, shared knowledge, and good escalation notes. No one knows every system, and many incidents require cooperation between help desk, infrastructure, security, vendors, and department managers. 10. Which ticketing systems have you used? If you have used a ticketing system before, name it and describe what you did with it: creating tickets, assigning priorities, documenting work, escalating issues, using canned responses, or building a small knowledge base. If you have not used one professionally, say so honestly, then explain that you understand the core workflow: capture the request, categorize it, prioritize it, document progress, communicate updates, and close the ticket when the user confirms the issue is resolved. Final Tips Before the Interview Practice explaining technical ideas in plain language. Prepare a few real examples from school, work, home labs, or previous customer service jobs. Do not pretend to know something you do not know. Explain how you would find out. When answering troubleshooting questions, start simple and move step by step. Show that you care about documentation, follow-up, and the user's experience. The strongest help desk candidates do not sound like they memorized a script. They sound like people who can stay calm, think clearly, and keep a user informed while working through the problem.
Short version: every AI feature we ship on the hosted helpdesk now runs on the self-hosted edition too. It's a separately licensed Docker add-on, perpetual license, one-time payment. Customer data stays on your network. Requires Jitbit Helpdesk v11.22 or newer. Why this took a while Our AI stack isn't a thin wrapper around an LLM API. It's a Python service running a local vector database (Qdrant), embedding models, and a RAG pipeline tuned against support-desk content. That's what makes "similar KB articles" surface the right article, and what keeps reply drafts grounded in your own documentation instead of hallucinated. Running that stack on our own hosting is one thing. Packaging it so your IT team can stand it up on your own hardware without babysitting Python dependencies is another. We sat on it until we had something we'd be comfortable supporting. What you get Similar-article suggestions inside tickets - ranked by semantic similarity, not keyword overlap Semantic KB search for end-users - results ranked by meaning AI-generated reply drafts grounded in your KB, writing-style rules, and the ticket context External documentation indexing - crawl your own docs, wiki, or any internal site and use it as AI context alongside the KB Choice of embedding model - free local model that runs on CPU, or OpenAI embeddings with your own key Generative provider - OpenAI, Azure OpenAI, or AWS Bedrock, customer brings their own key Feature parity with SaaS. No caveats about "basic" ChatGPT integration. How it ships A Docker Compose stack you drop in next to your existing Helpdesk install. Runs on Windows or Linux, bare metal or VM, Intel/AMD or ARM. Upgrades are zero-downtime and preserve Docker volumes — indexed data and cached models carry over. Full setup and system requirements are in the manual. Privacy This is the reason we finally built this. With the bundled local embedding model, nothing about your tickets, KB articles, or indexed documentation leaves your network. No embeddings sent to OpenAI. No vectors stored in a third-party service. The vector DB is a container on a host you own. If you want a generative model for reply drafts, you bring your own API key. Azure OpenAI and AWS Bedrock keep everything inside your existing cloud tenancy, with a BAA if you need one. For regulated on-prem buyers — healthcare, defense, financial services, government — this was the #1 reason you told us you couldn't adopt our AI features. It's no longer a reason. Pricing Licensed separately from the core Helpdesk product. Perpetual license, one-time payment, 1 year of updates included - same model as the rest of the on-prem lineup. No subscription, no per-agent add-on, no per-request fees. See the pricing page for the current price, and the on-prem AI landing page for the full feature list and requirements. Setup instructions live at jitbit.com/docs/ai-on-premise.
Do we even need helpdesk software? I mean, just give an AI agent a markdown skill file with your FAQ and canned responses and let it answer tickets - right? I (obviously) gave it a lot of thought recently and here's my take: If your entire support operation is one person answering the same three questions, congratulations, you've solved customer service. Ship it. For everyone else operating in reality - helpdesk apps do not automate support. They automate the messy human stuff around it. Workflow state & accountability Support isn't just answering questions - it's tracking who's answering them, who dropped the ball, and whose turn it is to care. Who owns this ticket? What's the SLA status? Did the second-line team even look at it, or did it rot in a queue for three days while everyone assumed someone else was on it? Teams need audit trails, escalation chains, and - let's be honest - blame-able history. An AI agent can triage, prioritize, categorize, and even draft a lovely response. It cannot enforce a process across a 70-person org where half the team is in a different timezone and the other half is "working from home" (AKA "at the beach"). Nobody's ripping out a tool for that just because ChatGPT can answer "how do I reset my password?" slightly faster. Customer data gravity You cannot put your entire docs website + a knowledge base into a markdown file. I mean, you can. You'd just need a web crawler to index your docs, dump them into a RAG database, and build an MCP server on top. Then maybe index the old tickets so the agent can search history, and... wait, you've just built a helpdesk app. With years of ticket history. Thousands of macros and canned responses. Customer sentiment patterns. Resolution time benchmarks. That one weird workaround for that one enterprise client that nobody remembers but the system does. That's institutional knowledge. That's training data. An AI agent starting from a markdown file has none of it. It's the new hire who didn't read the wiki - except the wiki doesn't exist yet either, because the wiki is the helpdesk history. Compliance & trust Enterprise buyers need GDPR compliance. Data residency. HIPAA. Audit logs that prove exactly who accessed what and when. "I built an AI agent over the weekend and pointed it at our support email" works great for a single-founder startup, but doesn't survive a procurement review. It barely survives a security questionnaire. Actually, it doesn't survive a security questionnaire - it is the security questionnaire's nightmare scenario. So what's the actual play I'm not here to dismiss AI - helpdesk apps need to absorb it. Embed AI deeply into the helpdesk itself: auto-drafted responses, smart routing, ticket summarization, triage, sentiment detection. Give customers the AI benefit inside the tool they already use, so they never feel the need to replace it. Better yet - become an MCP tool in your AI-powered org or even the orchestration layer. Let customers plug in their own AI agents, but manage them through the helpdesk. Routing rules, fallback-to-human thresholds, confidence scoring, handoff protocols. The helpdesk becomes the control plane, not the answer engine. Which, by the way, is exactly what we're building at Jitbit. The future isn't pure-AI support (customers will revolt) and it isn't pure-human support (too expensive). It's the helpdesk that best orchestrates humans and AI together. The markdown file crowd will figure that out eventually - right around the time their first enterprise prospect asks for an audit trail.
I've been dunking on AI pretty consistently on this blog. Not to mention I'm sick of all the AI influencer "built 15 apps in a weekend" crowd, but Claude Code just did something wild for me. Had it help us rip out our entire search engine - we're talking millions and millions of records, thousands of tenants - and migrate it from SQL Server full-text search to a small embedded, in-process Lucene port. Search went from 7-8 seconds down to milliseconds. Why now? For context - we relied on SQL Server full-text search for years and it was... mostly fine. The way a gas station sandwich is "mostly fine" when you're starving. It worked, nobody died, we had bigger problems to deal with. But when your customers are searching across millions of records and staring at a loading spinner for 7-8 seconds - "mostly fine" stops cutting it. I've been trying to approach this project for years and kept chickening out. Not because the search rewrite itself is hard - swapping a search provider is a weekend, maybe two. The thing that kept scaring me off was all the infrastructure around it. Dozens of one-off CLI tools for index rebuilding, compaction, deduplication, gradual rollout, health checks, verification scripts, progress reporting. Plus all the prep work - tech stack evaluation, risk analysis, benchmarking candidate libraries, planning a migration path that doesn't nuke search for thousands of paying customers at once. The kind of work that makes you reconsider your career choices. And this is exactly where AI came in. Not for the search code itself - I wrote that (well, most of it) in an hour. But for churning out dozens of self-contained CLI tools and scripts. Each one is a small, well-defined, green-field task. Clear inputs, clear outputs, no tangled legacy context to lose track of. If you read my jQuery rant - this was the exact opposite. That was brownfield hell, context rot after three files. This was "here's a spec, write me a tool," over and over, and Claude Code absolutely crushed it. And of course interrogating AI endlessly for subtle details like "Which folder is writable when you host ASP.NET Core in Docker? In Windows/IIS? In Linux?". How to keep RAM usage low for index writers? How to avoid locking issues when blue/green deployments overlaps? The final reindexing job that Claude helped me write ran for 39 hours straight across all tenant data, reporting nice progress graphs and auto-fixing errors as it went. It finished and everything checked out. Weeks of the most tedious infrastructure grind imaginable, compressed into days. Why Lucene For the uninitiated - Lucene is what powers Elasticsearch and Solr under the hood. Except we're not running an external Elasticsearch cluster. We're running a small custom fork of Lucene .NET port directly inside our app process. No extra services, no extra config, nothing. Stack Overflow did an almost identical migration back in 2011 (SQL Server full-text to Lucene.NET) and their reasons read like our internal planning doc: 1. Distribute the workload. Full-text search is heavy. With an embedded index, it happens right in the app process - no round-trip to the DB server, no waiting in line behind other queries. 2. Get the database off search duty. Our database is busy enough without demanding full-text queries piled on top. Pulling search out gives us headroom for actual SQL work - no more compromising between "what's good for full-text" and "what's good for everything else." 3. Better control over results. SQL Server full-text is a black box. Lucene gives you custom analyzers, field boosting, scoring tweaks - when a customer says "search isn't finding X," we can actually do something about it now. 4. No external service dependency. It's just code in our codebase, running in our process. No Elasticsearch cluster to provision, no separate infrastructure to monitor at 3 AM. A local folder for index files and that's it. 5. No new dependencies for self-hosted customers. A big chunk of Jitbit customers run our helpdesk on their own servers. "Hey, now you also need to set up and maintain an Elasticsearch cluster" - yeah, no. With an embedded library there's nothing new to install. Deploy the app, search works. Now, the tech porn For the developers in the room - here's what's actually running under the hood, because some of this was non-obvious. Index-per-tenant. Each customer gets their own Lucene index. No shared index with filtering - full isolation. One tenant's index compaction or reindex doesn't touch anyone else's. On SaaS we're talking thousands of indexes. On self-hosted, a single folder. LRU cache for IndexWriters. Lucene's IndexWriter is expensive to create - it acquires a write lock on the directory, loads segments, etc. Opening one per request would be murder. So we keep a pool of open writers in a custom LRU cache, capped at 50. When a writer gets evicted, the cache commits all changes and calls Dispose() on it automatically. In practice, 50 covers all active tenants with room to spare, and idle ones get quietly evicted. RAM buffer tuning. Each writer's RAM buffer is set to 2MB. That sounds stingy, but it's intentional. With potentially 50 live writers, you're looking at up to 100MB of indexing buffers just sitting there. Lucene defaults are much higher. We throttled it down and compensated with debounced commits instead. Debounced commits. We don't flush to disk on every document write - that would be slow and punishing on SSDs. Instead, every write schedules a debounced commit with a 5-second cooldown. If more writes come in, the timer resets. When things go quiet, it commits once. Batch writes during a reindex get explicit commits per chunk anyway. Blue-green deploy safety. Lucene uses file-level write locks. During a blue-green deployment, old and new instances briefly overlap. We set WriteLockTimeout to 5000ms - Lucene polls internally every second, so this gives the new instance five attempts to acquire the lock before giving up. Usually the old pod is gone by then. HTML sanitization before indexing. Ticket bodies come in as HTML. Indexing raw HTML means your search index is full of <div> tokens and CSS class names. We strip it with my own StripHTMLFast() (honestly I deserve a Nobel prize for this thing I wrote years ago, it uses Span<char> heavily and reads the HTML directly from the buffer stream without allocating any strings) before handing it to Lucene - both during reindex and on every incremental update. Sounds obvious, but easy to miss. Query escaping that doesn't break power users. We escape Lucene's special characters before parsing, but deliberately preserve " (phrase search), * (wildcard), and ? (single-char wildcard). So "exact phrase" and tick* both work. If the query still fails to parse after escaping, we fall back to wrapping the whole thing in quotes as a phrase search. Customers get power-user features without needing to know Lucene syntax. Resumable reindex. The reindex job writes a progress file after every 1000-ticket chunk, storing the last processed IssueID. If the server restarts mid-migration, it picks up from there. No starting over. On a 39-hour reindex across millions of records, that matters. Gradual rollout gate. Old instances (created before our migration cutover ID) don't auto-init Lucene - they fall back to SQL FTS until we explicitly triggered their migration. New instances above the threshold auto-trigger a background reindex on first search. This let us roll out to new signups immediately while migrating the legacy base in a controlled batch. OOM handling. If Lucene throws an OutOfMemoryException during a write, we catch it, evict the writer from the cache (freeing its RAM buffer), and rethrow. Better to lose in-flight writes than to leave a broken writer sitting in the pool corrupting future writes. So yeah Not gonna turn into one of those AI evangelists. But for the kind of tedious, well-defined infrastructure work that was blocking this migration for years - it saved me weeks, maybe months. Instead it took ONE F*CKING DAY and the 39 hours of staring at the autohealing reindexing job. The right tool for the right job and all that. Search is fast. Database is happy. Self-hosted customers don't need to install anything new. Ship it. Over and out, I'm off to enjoy the dopamine hit
More in programming
I listen to a lot of podcasts, and I like how they fit around other tasks. I press play, lock my phone, and put it down. I’m free to wash the dishes, fold the laundry, or shop for groceries. Unfortunately, more and more information is only published as a video. Technical talks, conference sessions, video essays – they don’t work in an audio-only podcast app. I could convert these videos to MP3 files, but that breaks down the moment a video isn’t pure spoken word. If a speaker says, “Look at this slide” or holds up a diagram, an audio-only file leaves me stranded. I don’t want to give up the podcast player I like, nor stare at a screen for an hour – but I do want the information in these videos. To solve this, I’m abusing my podcast player’s chapter support. This gives me the best of both worlds: I can listen to a video as audio-first, and glance at my lock screen if I need a moment of visual context. The idea: Chapters every few seconds MP3 files can have ID3 metadata, and ID3 metadata can include chapters. A chapter covers a particular time range, and it can have an associated title, description, and cover art. My podcast app of choice is Overcast, which can’t play videos, but it does have robust chapter support. I can jump between chapters, navigate a table of contents, and see per-chapter cover art. To get videos into Overcast, I’m creating MP3 files with a new chapter every few seconds, and the per-chapter cover art is a corresponding frame from the video. As I play the file, I get a slow, stop-motion-like rendition of the original video. If my phone is locked, I can glance at my lock screen and see the current frame in the Now Playing screen. Overcast is developed by Marco Arment, and I got this idea from Forecast, his app for adding chapters to podcasts. In particular, I was struck by its ability to create chapters that don’t display in the chapter list – ideal if I don’t want a table of contents with hundreds of entries. As I was developing my script, I compared my output to the output from Forecast to ensure I was creating the chapters correctly. The code: FFmpeg and Mutagen There are three steps in this process: Convert a video file to an MP3 Extract images from the video at a fixed interval Insert the images as hidden chapters in the MP3 file Let’s go through each in turn. 1. Convert a video file to an MP3 Converting a video file to an MP3 is a single FFmpeg command: ffmpeg -i video.mp4 audio.mp3 This is consistently the slowest step of the process, and I do wonder if I could use different settings or an alternative encoder to make it go faster – but it’s not slow enough to be worth further investigation. 2. Extract images from the video at a fixed interval Extracting images from a video needs a more complicated FFmpeg command: ffmpeg -i video.mp4 \ -vf 'fps=1/5,scale=iw*sar:ih,scale=min(iw\,945):min(ih\,945):force_original_aspect_ratio=decrease' \ thumbnail_%04d.jpg This extracts an image every 5 seconds, downscales any image larger than 945 pixels square (while preserving the original aspect ratio), and saves the results as sequentially numbered JPEG images (thumbnail_0001.png, thumbnail_0002.png, and so on). The key is the -vf flag, which defines two FFmpeg filters: The fps filter selects one frame every 5 seconds (fps=1/5). The first scale filter scales the width based on the sample aspect ratio (scale=iw*sar:ih). Without this filter, frames can be stretched and distorted. The second scale filter scales the input video, preserving the original aspect ratio (force_original_aspect_ratio=decrease), and ensuring the output images fit within 945×945px or the size of the input video, whichever is smaller. My limit is 945 pixels because that’s the largest size that cover art is shown on my iPhone. This filter still isn’t completely correct – it sometimes creates images from portrait videos that are smaller than I’m expecting – but it’s good enough. These are only thumbnails for glancing at, and if I want to change it later, I can always do the image resizing outside FFmpeg. 3. Insert the images as hidden chapters in the MP3 file Inserting the chapters into the MP3 file is more complicated. Although FFmpeg has basic support for ID3 metadata, as far as I know, it can’t insert chapters with per-chapter artwork. Instead, I’m going to reach for Python and the Mutagen library. Here’s the code to add a chapter to an MP3 file: from mutagen.id3 import APIC, CHAP, ID3, PictureType audio = ID3("audio.mp3") with open("thumbnail_0001.jpg", "rb") as f: img_data = f.read() image_frame = APIC(mime="image/jpeg", type=PictureType.OTHER, data=img_data) chapter_frame = CHAP( element_id="chp1", start_time=0, end_time=5 * 1000, sub_frames=[image_frame] ) audio.add(chapter_frame) audio.save() This creates a single chapter that lasts the first 5 seconds (0 to 5000 milliseconds), and the per-chapter cover art is thumbnail_0001.jpg. If we ran this in a loop, we could add images for every 5 second slice of the original video. This code is inserting two frames into the ID3 metadata: The CHAP (chapter) frame contains the timing information, and it can have subframes for metadata like title, chapter art, or associated URL. The APIC (attached picture) subframe contains information about a picture, which can either be a blob of image data or a URL to an image on the web. Normally, you’d also insert a CTOC frame which defines a table of contents, but I don’t want a TOC with hundreds of 5-second chapters, so I’m deliberately not doing this here. This is allowed by the ID3 spec – you’re not required to insert a CTOC frame if you’re using chapters, and you can have chapters that aren’t listed in your table of contents. To work out which frames I needed, I used Forecast to create some chapters by hand, and I inspected their frames. In particular, loading an MP3 and calling Mutagen’s pprint() method shows a human-readable list of frames, and then I could drill into the individual fields: from mutagen.id3 import ID3 audio = ID3("audio.mp3") print(audio.pprint()) I wrapped all this code in a project called glancecast, which allows you to convert a video file with a single command, with optional flags to set the frame length and chapter art size: $ python3 glancecast.py interesting_talk.mp4 interesting_talk.mp3 The process takes a minute or so to complete, most of which is spent transcoding the video file to MP3. The resulting MP3s are usually 40 to 50 MB in size, which is very reasonable. The outcome: How it looks in practice Here’s what one of these “glanceable” podcasts looks like in Overcast and on my lock screen: Maggie Appleton presented this talk over two years ago and it’s been on my “talks to watch” list ever since. Once I put it in Overcast? I listened to it in less than a day. It’s not a lot of extra information, but enough that I can quickly glance down and get the gist of what a speaker is saying. Both views update with a new frame every few seconds, or I can put my phone in my pocket and ignore the screen. I’ve used this approach for half a dozen videos so far, and I’m happy with the results. I expect to keep using it, because I have a long queue of videos I’ve been meaning to watch. If you’d like to try this, check out glancecast for the full code and instructions. [If the formatting of this post looks odd in your feed reader, visit the original article]
Andrew Baker, the current Group CIO at Capitec Bank wrote an interesting piece on AI and open source, and how these tools that generate code according to one’s specification may replace the general reliance on open source implementations done by contributors around the world. I’d really recommend reading it. I have great admiration and respectContinue reading "AI Isn’t Replacing Open Source"
A framework for thinking about when AI involvement is additive or a violation
Why we need richer, thicker interfaces and better boundary objects for collaborative planning with agents
A look at 10 foundational pillars that enable agents to operate more competently and more efficiently in any codebase.