As expected. Almost the whole company is gone, less than 15 people left in engineering.

Tell HN: Bending Spoons laid off almost everybody at Vimeo yesterday

Hi,I started Muky in April 2024. Classic side project that got out of hand. We have two kids - the younger one is happy with the Toniebox, but our older one outgrew it. She started asking for specific songs, audiobooks that aren&#x27;t available as figurines, and &quot;the music from that movie.&quot;We had an old iPad Mini lying around and already pay for Apple Music. Felt dumb to keep buying €17&#x2F;$20 figurines for 30-45 minutes of content when we have 100 million songs.Now at version 4.0 after ~20 updates. Some lessons:On the hardware vs app tradeoff:
Toniebox and Yoto are brilliant for little ones – tactile, simple, no screen needed. But they hit a wall once kids want more. And handing a 5-year-old Apple Music means infinite scrolling and &quot;Dad, what&#x27;s this song about?&quot; Muky sits in between – full library access, but parents control what&#x27;s visible.On sharing:
Remember lending CDs or cassettes to friends? Or kids swapping Tonie figurines at a playdate? I wanted that for a digital app. So I built QR code sharing. Scan, import, done. And unlike a physical thing – both keep a copy.On onboarding:
First versions: empty app, figure it out yourself. Retention was awful. Now: 4-step onboarding that actually guides you. Should&#x27;ve done this from the start.On content discovery:
100 million songs sounds great until you have to find something. Parents don&#x27;t want to search – they want suggestions. Spent a lot of time building a Browse tab with curated albums and audiobooks for kids. Finally feels like the app helps you instead of just waiting for input.On going native:
Went with Swift&#x2F;SwiftUI instead of Flutter or React Native. No regrets - SwiftUI is a joy to work with and performance is great. Android users ask for a port regularly. No capacity for that now, but Swift for Android is progressing (<a href="https:&#x2F;&#x2F;www.swift.org&#x2F;documentation&#x2F;articles&#x2F;swift-sdk-for-android-getting-started.html" rel="nofollow">https:&#x2F;&#x2F;www.swift.org&#x2F;documentation&#x2F;articles&#x2F;swift-sdk-for-a...</a>). Maybe one day. CarPlay is another one parents keep asking for – going native should make that easier to add, if Apple grants me the entitlement.On subscriptions vs one-time:
Started with one-time purchase. Revenue spikes at launch, then nothing. Switched to subscription – existing one-time buyers kept full access. Harder to sell, but sustainable.Ask me anything about indie iOS dev or building for kids. App is at <a href="https:&#x2F;&#x2F;muky.app" rel="nofollow">https:&#x2F;&#x2F;muky.app</a> if you&#x27;re curious.

Tell HN: 2 years building a kids audio app as a solo dev – lessons learned

I&#x27;ve been trying to get agentic coding to work, but the dissonance between what I&#x27;m seeing online and what I&#x27;m able to achieve is doing my head in.Is there real evidence, beyond hype, that agentic coding produces net-positive results? If any of you have actually got it to work, could you share (in detail) how you did it?By &quot;getting it to work&quot; I mean:
 * creating more value than technical debt, and
 * producing code that’s structurally sound enough for someone responsible for the architecture to sign off on.Lately I’ve seen a push toward minimal or nonexistent code review, with the claim that we should move from “validating architecture” to “validating behavior.” In practice, this seems to mean: don’t look at the code; if tests and CI pass, ship it. I can’t see how this holds up long-term. My expectation is that you end up with &quot;spaghetti&quot; code that works on the happy path but accumulates subtle, hard-to-debug failures over time.When I tried using Codex on my existing codebases, with or without guardrails, half of my time went into fixing the subtle mistakes it made or the duplication it introduced.Last weekend I tried building an iOS app for pet feeding reminders from scratch. I instructed Codex to research and propose an architectural blueprint for SwiftUI first. Then, I worked with it to write a spec describing what should be implemented and how.The first implementation pass was surprisingly good, although it had a number of bugs. Things went downhill fast, however. I spent the rest of my weekend getting Codex to make things work, fix bugs without introducing new ones, and research best practices instead of making stuff up. Although I made it record new guidelines and guardrails as I found them, things didn&#x27;t improve. In the end I just gave up.I personally can&#x27;t accept shipping unreviewed code. It feels wrong. The product has to work, but the code must also be high-quality.

Ask HN: Do you have any evidence that agentic coding works?

Status page is up, but can&#x27;t seem to git pull or push

Ask HN: Is GitHub Down?

Seeing a lot of people running coding agents (Claude Code, etc.) in custom sandboxes Docker&#x2F;VMs, firejail&#x2F;bubblewrap, scripts that gate file or network access.Curious to know what&#x27;s missing that makes people DIY this? And what would a &quot;good enough&quot; standard look like?

Ask HN: Why are so many rolling out their own AI/LLM agent sandboxing solution?

Looking at the https:&#x2F;&#x2F;www.githubstatus.com&#x2F;history there&#x27;s been a significant issue in GitHub 13 out of 22 days this year. These track issues that warrant their own status page, but from personal experience there&#x27;s some sort of degraded performance to be seen in the service almost daily, most often in the evening in my local time.As a long time paying GitHub user the stability has been bad for years now, but this has been an upwards trend that&#x27;s only been accelerated in the recent times. What&#x27;s happening inside of GitHub and can we expect it to ever get better anymore?

Tell HN: GitHub has experienced issues 60% of days this year

It&#x27;s 2026 and I shouldn&#x27;t spend 3 hours every month manually splitting $15K revenue:• 50% to co-founder
• 10% across 3 contractors
• 5-10% to ~15 affiliates
• 30% to tax accountThis should be automated. Maybe through a kind of workflow builder that can trigger money flows.What I&#x27;ve tried:- Stripe Connect: Only splits to one account
- Zapier: Can&#x27;t actually move money (ToS restriction)
- Manual scripts: Works but I&#x27;m now maintaining financial infrastructure. 
- Escrow.com: $100 min fee, designed for one-off transactionsWhat I want: Set rules once → money routes automatically each month.Questions:1. Does this exist? (Feels like I&#x27;m bad at searching)2. If not, why? Regulatory? Nobody trusts automation with money? Technical blocker? Stable coins could maybe help shipping this.3. What&#x27;s your current solution? Custom code? Just manual transfers?I&#x27;ve talked to ~20 founders. Most are either:
- Writing custom scripts (requires dev skills + maintenance)
- Paying accountants (expensive, still manual)
- Suffering through manual transfers (time sink)Seems like a gap between &quot;fully manual&quot; and &quot;build your own payment infrastructure.&quot;Am I missing something obvious?

Ask HN: Does "Zapier for payment automation" exist?

With the increase of vibe coding I am interested in knowing some creative ways people are automating their coding work.

Ask HN: How are you automating your coding work?

I am a max subscriber for claude.ai using their browser UI and claude desktop ai app, not a API user, claude code user. In the last week I have come across every session of mine not lasting more than an hour or so. Have any of you come across this? Is this something new? Reaching out to their help is basically dealing with claude and canned responses, escalated to human but they&#x27;ve yet to explain&#x2F;respond

Tell HN: Claude session limits getting small

Curious to hear from anyone actively working with COBOL&#x2F;mainframes. Do you see LLMs as a threat to your job security, or the opposite?I feel that the mass of code that actually runs the economy is remarkably untouched by AI coding agents.

Ask HN: COBOL devs, how are AI coding affecting your work?

Hello :-)I have a Discord server I set up a long time ago. Around 2016 I think. Back then, it was lively and active and loads of fun. Over time it&#x27;s developed close to 5,000 members (it actually had over 5,000 members at one point) and currently has 501 members online as I type this. It&#x27;s more likely there&#x27;s about 10-15 that are paying attention to anything happening.It&#x27;s a Discord that originally focused on DevOps. It complemented my YouTube channel on the same topic, but since then, as it&#x27;s slowly died out, and my channel&#x27;s focus as shifted and changed, it&#x27;s become a bit of a waste land.It&#x27;s a shame really, because a really fun Discord server can be a great place to be, but I&#x27;m not sure where to take it now.How would you handle this situation? What would be your approach to reviving the Discord and perhaps trying to get a community of like minded hackers going again in 2026?I won&#x27;t link the Discord here as I&#x27;m not trying to beg for users or spam. I just genuinely want to work on a solution to improve the life of the server. I will put it in my HN profile, though, so if you do want to check it out that extra step is required.Are people even interested in Discord servers any more? I don&#x27;t know.Thanks in advance.

Ask HN: Revive a mostly dead Discord server

Anthropic has these docs that use git worktree: <a href="https:&#x2F;&#x2F;code.claude.com&#x2F;docs&#x2F;en&#x2F;common-workflows#run-parallel-claude-code-sessions-with-git-worktrees" rel="nofollow">https:&#x2F;&#x2F;code.claude.com&#x2F;docs&#x2F;en&#x2F;common-workflows#run-paralle...</a>There are some apps like this that leverage git-worktrees under the hood: <a href="https:&#x2F;&#x2F;conductor.build">https:&#x2F;&#x2F;conductor.build</a>I&#x27;ve tried lazygit as well to make it more convenient.I still end up preferring having multiple clones of my repo when I need to ensure the agents don&#x27;t accidentally overlap.Curious what other people do.

Ask HN: How do you run parallel agent sessions?

Since upgrading to 15.7.3 on my main machine (M4 Pro mini), every time I start I get asked if I want to enable automatic updates or just download only. Regardless the response I then get a welcome message and it happens again at the next boot.I have installed the latest Tahoe on my 2nd machine (M1 mini) and it&#x27;s complete dog shit. I refuse to subject myself to that until they fix some of the egregious problems (as documented numerous times on this forum).Apple seem to be desperate to alienate their userbase.

Ask HN: Does Apple not have *any* QA for their older macOS releases?

I&#x27;m trying not to get too caught up in the hype here. I&#x27;m curious what you&#x27;ve actually built or shipped so far with claude-code.I made a little phonics flashcard game for my kids <a href="https:&#x2F;&#x2F;apps.apple.com&#x2F;us&#x2F;app&#x2F;3letterstories&#x2F;id6753956099">https:&#x2F;&#x2F;apps.apple.com&#x2F;us&#x2F;app&#x2F;3letterstories&#x2F;id6753956099</a>Also related to that: a tool to fine tune the images on the flash cards (which are all AI generated on gemini).Some internal tooling - a self contained page that does JSON&#x2F;python dict visualization - this obviously exists already.Nothing big yet, but it&#x27;s only been a couple of days. This tool does seem really good at building frontends&#x2F;dashboards.

Ask HN: What have you built/shipped with Claude Code?

I have a working theory: If users think your app is complex, it does not mean they hate it, they just have higher expectations&#x2F;reward for putting up with it.I have worked on making software easy for users, and often I got them to see the app has no value much faster because by stripping complexity, I lost the tiny functionalities that made it useful for them.

Users don't care about your app's complexity

I was an Enterprise customer on their platform. Cerebras began terminating production models and replacing them every few months. This time they notified us they are terminating Llama 3.3 70B, which is the model my plan was subscribed to.Instead of offering us an alternative plan they told us all others are sold out and have kicked us off their platform.Their Discord support group has multiple Enterprise customers all getting the same treatment. Their own support staff is telling them to migrate to Groq.You can’t build a stable business with a company that randomly terminates models this frequently. Who does not respect their customers (especially Enterprise customers, not even small utilizers), offering zero contingency plans if they decide to axe your model.Because of their architecture, they can only host a finite number of models. Turnover is fast, so if your model gets marked for deprecation, you will get forced off the platform.Cerebras is cool for personal Hobby projects, but is in absolutely no position to be selling &quot;Enterprise&quot; accounts to businesses.

Tell HN: Avoid Cerebras if you are a founder

Hi HN,I’ve been experimenting with letting LLMs generate and then continuously modify small business applications (CRUD, dashboards, workflows). The first generation usually works — the problems start on the second or third iteration.Some recurring failure modes I keep seeing:
 • schema drift that silently breaks dashboards
 • metrics changing meaning across iterations
 • UI components querying data in incompatible ways
 • AI fixing something locally while violating global invariantsWhat’s striking is that most AI app builders treat generation as a one-shot problem, while real applications are long-lived systems that need to evolve safely.The direction I’m exploring is treating the application as a runtime model rather than generated code:
 • the app is defined by a structured, versioned JSON&#x2F;DSL (entities, relationships, metrics, workflows)
 • every AI-proposed change is validated by the backend before execution
 • UI components bind to semantic concepts (metrics, datasets), not raw queries
 • AI proposes structure; the runtime enforces consistencyConceptually this feels closer to how Kubernetes treats infrastructure, or how semantic layers work in analytics — but applied to full applications rather than reporting.I’m curious:
 • Has anyone here explored similar patterns?
 • Are there established approaches to controlling AI-driven schema evolution?
 • Do you think semantic layers belong inside the application runtime, or should they remain analytics-only?Not pitching anything — genuinely trying to understand how others are approaching AI + long-lived application state.Thanks.

How do you keep AI-generated applications consistent as they evolve over time?

I&#x27;ve been working as a Software Engineer for 20+ years.Places I worked in the early years barely had an IT department at all. As a developer you were expected to be able to maintain your machine. We&#x27;d install whatever we want, experiment with different operating systems, etc. Total free rein, box was our tool to get work done with, they didn&#x27;t care how you did it.That went away a long time ago. Basic corporate spyware and rules came pretty early but still free rein over our tools.I&#x27;ve worked with the same company for close to a decade now, and they have been tightening and tightening the noose slowly but surely. We&#x27;re purportedly a software company, but we lost admin rights, installable software went from a blocklist to an allowlist. Everything we install needs to get approved by IT, and that approval takes weeks.Today they took our Chrome extensions away. They&#x27;ve got an allowlist of about 15 extensions we can install. Everything I submitted for approval got rejected.I&#x27;m frustrated with this arrangement and am wondering how standard this is these days in this industry?So I&#x27;m genuinely curious, Hacker News: How big of a company do you work for, what industry, and how locked down is your machine?

Ask HN: How locked down are your work machines?

What happened?Gemini 12 months ago was dead last the worst of all LLMS.Then it streaked miles ahead of the others to be absolutely unbelievably good.Then a month or so ago it turned back into being garbage. I waited for it to be fixed but it has remained bad.I do not understand.

Ask HN: How did Gemini go from being awful to incredible back to awful?

I’ve been using GitHub Copilot’s free version for inline suggestions and occasional tasks at work. I’ve also heard a lot of hype around tools like Cursor, Claude Code, and OpenAI Codex, but I haven’t had time to dive into them yet.Now that I have some free time, I’m looking for high-quality resources that:1. Give a good overview of the current AI code editor landscape2. Explain tool-specific features (e.g., recent features like skills, etc.)3. Provide tactical tips for using these tools effectively in real development workflowsBlog posts, videos, tutorials, beginner to advanced guides, or even personal workflow tips would all be helpful.

Ask HN: What are good resources to get familiar with AI code editors?

I’m trying to understand what happens when non-technical teams build AI apps through vibe coding, and those apps later become widely used inside a company.How do teams then handle monitoring, bugs, scaling, and reliability with many users?

Ask HN: Vibe-coded prototypes: what happens when they go into production?

I sell 3d printed sleeves to attach an Apple Airtag to a Samsung TV remote. But Amazon thinks I am selling a Samsung device and has deactivated me for IP violations. 
My listings are very clear that this is FOR a Samsung device, and not an actual Samsung device. But Amazon&#x27;s automated system can&#x27;t figure it out. I am following their IP guidelines for compatible products (section 6C - https:&#x2F;&#x2F;sellercentral.amazon.com&#x2F;help&#x2F;hub&#x2F;reference&#x2F;GZUQ6GBBXQVHQKF2):You can see one of my listings here:https:&#x2F;&#x2F;www.amazon.com&#x2F;dp&#x2F;B0FHZVVL1HEvery time I listed a new product for a different TV remote, my item would be flagged, but then I would go in and make a small edit, and that seemed to trigger some sort of review, and everything would be great.Until last week, out of nowhere, my account has been deactivated.To reinstate my account, they asked me to submit an authorization letter from the manufacturer or brand owner authorizing me to sell their products. Well, I figured, I am the manufacturer, so I wrote an authorization letter for myself, and even got it notarized for good measure, that I am legally allowed to sell these devices. But to no avail.I have an option to &quot;Submit new information.&quot; But have no new information to submit, and fear if I try submitting anything else, I&#x27;ll be permanently banned or something.The funny part is that on most of the products I have listed, I am losing money, just because of the FBA costs and the advertising costs. I lost about $250 last month between two of the variants.The sad part is that I sell a book, Computer Engineering for Babies, and do most all my sales through my website, but do get a few orders a week through amazon for the book, and am now afraid the Amazon door is closed forever.

Tell HN: Amazon has deactivated my seller account

I&#x27;m probably a bit late to the party. But I just want to tell my short story, because I&#x27;m amazed. I have several open source project and especially one that I use every day. It&#x27;s not a project with huge user-base, but there are some. Some of the user has posted issues, which I have wanted to take care of, but time, is always an issue. Debugging, trying to get into issues other users expirence while I don&#x27;t, is always harder to get grip of.So, today I fired up Claude Code, told it about my project and during 2h (including manual review of the generated code), I&#x27;ve fixed several issues. I even managed to get in new features I have wanted as well!<a href="https:&#x2F;&#x2F;github.com&#x2F;Lallassu&#x2F;gorss" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Lallassu&#x2F;gorss</a> is now v0.5 thanks to this.

Tell HN: Claude helped me maintain my old open source project

I see an urgent need for a map projection that makes Greenland look as small as possible.  What are the options?

Ask HN: Which common map projections make Greenland look smaller?

I&#x27;m exploring building a JavaScript obfuscator and trying to understand if there&#x27;s real demand — from indie devs to enterprise.The reason: every web app ships source code to the browser — always been analyzable, copyable, patchable. Now with AI able to explain minified code in seconds, I&#x27;m wondering if attitudes are shifting.If you own any web apps or games, do you genuinely care about protecting client-side code? If not, why? If yes, what do you use — or why don&#x27;t existing tools work for you?(Context: I&#x27;m building www.afterpack.dev — but genuinely want to understand if this is a problem people actually have or if it&#x27;s rather a very niche problem)

Ask HN: Do you protect your client-side JavaScript? Why or why not?

I’ve been experimenting with a compiled language design and wanted feedback from people who’ve worked on compilers or systems code.The core idea is allowing multiple execution profiles to coexist in a single source file:- userland (default): indentation-based, Python-like syntax, safety by default
- kernel: brace-based syntax, strict rules, no heap or runtime
- baremetal: brace-based syntax, raw pointers, no safety netAt build time, a single profile is selected and all other code is erased at compile time — no runtime checks or overhead.The compiler pipeline is lexer → parser → typed IR → LLVM backend, with profile rules enforced during IR validation rather than at runtime.Roughly ~90% of the core language and compiler are implemented; I plan to open-source the project once the remaining pieces are finished.This is still an experiment. I’m mainly curious:
- does this model make sense?
- are there obvious design pitfalls?
- has anyone seen similar approaches succeed or fail?I’d appreciate critical feedback.

Experimenting with a Compiled Language

Do they ever ask themselves why they can&#x27;t search for an exact string?Do they ever ask themselves why they can&#x27;t filter out little signature logos when searching for emails that have attachments?thank you for any sincere responses

Do people at Google use Gmail?

For those that used to attend in-person technical events before 2020, are you going again?It feels to me like the US is slower to return to in-person events than other parts of the world.

Ask HN: Are you going to meetups/conferences?

Context: I’m working on a compliance preparation tool for early-stage startups, and I’ve spoken with many teams going through SOC 2 &#x2F; ISO 27001. I’m posting here to sanity-check my understanding and learn what others found most painful before the audit.
Most teams don’t delay SOC 2 because they don’t care about security or because customers aren’t asking. They delay because it’s extremely unclear how to start.You Google “SOC 2” and you’re immediately hit with:
- 100+ controls
- Type I vs Type II
- Trust Services Criteria
- Tooling vs auditors vs consultants
- The result is that many startups treat SOC 2 as a tooling problem.They wait until a deal is blocked, then:
- Sign up for Vanta or Drata
- Hire a consultant
- Try to “speedrun” complianceWhat actually hurts them isn’t missing controls — it’s missing readiness. No clear asset inventory, no ownership, no risk model, no vendor tracking, no idea what evidence even exists yet.By the time tools or auditors enter the picture, everything is reactive and expensive.For those of you who’ve been through SOC 2:
- What helped you most before the audit?
- What do you wish you had done 3–6 months earlier?
- Did you start with tools, docs, or internal processes first?Genuinely curious how others approached this.

Ask HN: Why does SOC 2 feel so hard for early-stage startups?

Building agents that can spend money (via AP2&#x2F;payment APIs). Regulators want audit trails. Agents use tools from multiple vendors.Problem: No unified way to prove why an agent made a decision. Each vendor logs their piece, but the full decision chain is fragmented.Anyone solved this? Especially interested in: 
- Cross-vendor decision provenance 
- What to log (CoT? Context? Just actions?) 
- Storage (centralized DB vs. immutable ledger)

Ask