I just spent 1h+ debugging why my locally-hosted gitlab runner would fail to create pipelines. The gitlab job output would just display weird TLS errors when trying to pull a docker images. After debugging gitlab and the runner, I realized after a while I could not even run &quot;docker pull &lt;image&gt;&quot; on my machine as root:&gt; error pulling image configuration: download failed after attempts=6: tls: failed to verify certificate: x509: certificate is not valid for any names, but wanted to match docker-images-prod.6aa30f8b08e16409b46e0173d6de2f56.r2.cloudflarestorage.comFirst blaming tailscale, dns configuration and all other stuff. Until I just copied that above URL into my browser on my laptop, and received a website banner:&gt; El acceso a la presente dirección IP ha sido bloqueado en cumplimiento de lo dispuesto en la Sentencia de 18 de diciembre de 2024, dictada por el Juzgado de lo Mercantil nº 6 de Barcelona en el marco del procedimiento ordinario (Materia mercantil art. 249.1.4)-1005&#x2F;2024-H instado por la Liga Nacional de Fútbol Profesional y por Telefónica Audiovisual Digital, S.L.U.
https:&#x2F;&#x2F;www.laliga.com&#x2F;noticias&#x2F;nota-informativa-en-relacion-con-el-bloqueo-de-ips-durante-las-ultimas-jornadas-de-laliga-ea-sports-vinculadas-a-las-practicas-ilegales-de-cloudflareFor those non-spanish speakers: It means there is football match on, and during that time that specific host is blocked. This is just plain madness. I guess that means my gitlab pipelines will not run when football is on. Thank you, Spain.

Tell HN: docker pull fails in spain due to football cloudflare block

What are you working on?  Any new ideas that you&#x27;re thinking about?

Ask HN: What Are You Working On? (April 2026)

Here&#x27;s hoping that it will return soon, as I really liked it.

Tell HN: OpenAI silently removed Study Mode from ChatGPT

Hey folks,I&#x27;m building an Open Source email newsletter tool and one of the challenges we have is form spam: As soon as a signup form goes live somewhere, bots will try to sign up. This is possibly an attempt to overwhelm the inbox of people whose accounts have been compromised. But it&#x27;s also bad for the people who run the newsletter as these ultimately unwanted emails reduce their sender reputation.There was recently a discussion here on HN about this topic [1]. The post author ended up using Cloudflare Turnstile to mitigate the issue. We currently already have support for external captcha solutions like hCaptcha. However, many of our users are quite privacy-conscious and don&#x27;t like having user data sent to third parties (especially non-EU third parties for our European users).So now I&#x27;ve been thinking of adding an invisible proof-of-work (PoW) captcha to all signup forms. Possible implementations I&#x27;ve been considering are Altcha [2] and mCaptcha [3].Now to my question: Have any of you tried using PoW captchas to protect against form spam? What have your experiences been with it so far?[1] https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=47609882[2] https:&#x2F;&#x2F;altcha.org&#x2F;[3] https:&#x2F;&#x2F;mcaptcha.org&#x2F;

Ask HN: What's your experience with PoW captchas against form spam?

I was writing a comment recently when I realized just how bad the graphs in GPT 5 video are. I had almost forgotten about it.I wish to create a very minor website which can talk about all of these, I feel like there are some perverse incentives which make these companies show out the good and hide out the bad and our news cycle just keeps on moving and we all follow it.Also there can be some remarkable mentions as well which I wish to listHere are some other examples:OpenAI turning from non profit to open.Claude code software getting leaked (Anthropic)X&#x2F;Grok generating non consensual deepfakes issue (XAI)OpenAI&#x27;s deal with Military and backstabbing basically and now pushing for new laws.The graph thing by OpenAI being so horrendous which I mentioned earlier.Suchir balaji&#x27;s death (I genuinely recommend searching this up if someone hasn&#x27;t for context)Microsoft Co-pilot thingsAnthropic when detected that chinese models distill from them showed that they can view things sent in API and there were some comments related to it which I forgot but it doesn&#x27;t sound privacy friendly.The stargate project by OpenAI,Oracle etc. leading to the ram crisis in the first place.Remarkable mentions list includes:OpenAI&#x27;s Sam altman visible stumbled and saying now iconic line ‘Want to sell your shares? I’ll find you a buyer&#x27;People mentioning how these AI companies have reduced the amount of messages people can send and also all the comments about the dumbing down of these models while still charging 200$ 100$ for these things.Microsoft has named 75 products I think all on Co-pilotAnthropic treating their mythos model like their own baby in a very psychotic way.Deepseek giving you answers in high priority in their website when it was first released if you said that Xi Jinping wanted the work (speculation-ish)I would love if people can give more and more examples about it and I wish to compile a list and have a discourse surrounding this. Thanks for reading this and have a nice day and waiting to read the comments here!

Ask HN: What are all the bad things that AI companies have done which we forgot

are you ok sharing secrets or api keys to you ai agent via .env?or is there any other tool or mechanism that one use to safegaurd from potential exploit or leaks

Ask HN: Do you trust AI agents with API keys / private keys?

I think Nostr is so much interesting and powerful to create stuff for free.

Ask HN: Anyone using Nostr as a lightweight back end/DB for rapid prototyping?

I&#x27;m looking for something novel and interesting, that isn&#x27;t absolutely crowded that I could meaningfully contribute to.In 2022 I was toying around with OpenAI&#x27;s RL Gym, right when the first non-instruct GPT3 model came out. I was thinking about getting into ML a lot more, but hesitated.
Before that it was 3D printers, mechanical keyboards, drones, etc.
All of these have exploded, and while they are still very interesting, I do love my Browns and manage Prusas for my local hackerspace, they have just, for the lack of a better term, industrialized.
I&#x27;m also now in a position where I have time and money for it, not like when I was 15 and rating Ender motherboard upgrades I knew I&#x27;d never buy.Right now I&#x27;m making a chess engine, but that&#x27;s already a solved problem.
There&#x27;s also biohacking, and while designing chips to go into my body is really interesting, I only have one, and don&#x27;t want to push it too far.
One promising idea is a kind of &#x27;Personal Computer 2&#x27;, where people try to innovate HCI, and while I really like that and do have some research ideas, I&#x27;d like to explore a bit more before delving deep into it.

Ask HN: Any interesting niche hobbies?

Hey guys, 
My app PairCare has 2-9 downloads every single day but haven&#x27;t had new trials for weeks so far. 
It&#x27;s been a month after launch, a total downlaods is around 130 but I got 3 real subscribers and 3 people tried and canceled right away. Only one tried and canceled when it&#x27;s almost to subscribe. I&#x27;m thinking it&#x27;s because of onboarding so I&#x27;m gonna fix it but maybe it&#x27;s time to give up... I don&#x27;t know. If you have any advice I&#x27;ll appreciate it.

Ask HN: What should I do with my app? 130 downloads 3 real subscribers

I&#x27;ve reviewed hundreds of decks as an early-stage fund manager. The decks have 
gotten better every year. The businesses underneath them haven&#x27;t kept pace.My observation: founders optimize for the 5-minute impression. Investors evaluate 
for the 5-year business. The pitch coaching industry selects for performance over 
clarity. The best decks I&#x27;ve seen are often less polished — but they answer every 
investor question before it&#x27;s asked.Curious whether others — founders, investors, operators — see the same pattern. 
Is the deck a useful forcing function or a distraction from the actual work?I wrote about this here if useful context:
[https:&#x2F;&#x2F;www.pitchvault.ai&#x2F;pitch-intelligence&#x2F;pitch-deck-performance-vs-business]

Is the pitch deck culture making founders worse at building businesses?

I don&#x27;t have anything against AI, but HN (and everywhere else) seems to be drowning in AI atm.Seems like every man and his dog is building an AI agent harness. And power to you (and your dog) if that&#x27;s you.But it would be refreshing to hear about some non AI related projects people are working on.

Ask HN: What are you building that's not AI related?

I saw the HackerRank (YC S11) hiring post (https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=47667011) and it made me realize I no longer understand how to evaluate candidates effectively.Specifically, we are changing hiring across 3 dimensions:
&gt; Tasks: Real-world tasks on code repositories vs standard algorithmic-style puzzles
&gt; Evaluation: AI fluency, orchestration skills vs functional correctness
&gt; Candidate experience: Agentic IDE vs a simple code editorIn the “old world,” you could ask multiple questions and triangulate skill from answers. Now it seems like evaluation depends heavily on tools and models that keep changing month to month.So I’m curious:
&gt; What signals actually correlate with strong engineers today?
&gt; How do you design interviews that don’t become obsolete with the next model release?
&gt; Are algorithmic interviews still useful at all?Would love to hear from people who have recently changed their hiring process or have been interviewed using this new approach.

Ask HN: Hiring in the age of AI-assisted coding: what works?

I have this long thought about building a language for the age of ai.Here is a non-exhaustive list of features i would love to have it included:
- separate effect full code from pure logic
- separate state and code
- ai native debugger
- native deterministic simulation testing
- coding proofs
- explicit type system
- repl
- wasm targetI have no background in compiler building or similar. I would just like to spend some weeks reading the best resource about this topic.Please share any resource you recommend reading.

Ask HN: Best books on building a programming language

I get good, cheap, fast feature coding success with grok-4.1-fast for planning and grok-code-fast-1 for execution. But according to the Openrouter usage stats, grok-code-fast-1 is now old hat - usage dropped off a cliff in mid-Feb.What are you using now? (That&#x27;s around the same price per token) (AND also blazing fast (100+tps)? is that possible?)(I&#x27;m finding Minimax M2.5 usable, but I think I still prefer grok-code-fast-1. Opus rocks but is expensive - good to have in the quiver for planning a big refactor.)

Ask HN: Former grok-code-fast-1 users, what coding model are you using now?

Given a system with N components and M pairwise connections between said N components, there exists infinite sequences of [component, command, ...] for a series of operations. A component displays data and a command obviously executes a command.Has anyone here done research on using an agant to select happy paths to test based on user intent? Is there literature on this? Seems like a cost, compute, intent optimization problem.Assume latency is a restriction.Seems like something some AI research folks here can chime in on. I have my own opinions but want to hear what the community thinks.

Ask HN: Agentic Permutation of Testing Paths In A System

I&#x27;ve shipped multiple products over the past few years. Every single one followed the same pattern: build, post, get 12 likes from friends, a bit of organic traction, then nothing. Back to coding a new thing.I know I need marketing help but giving equity to someone I met online feels like a huge risk. At the same time hiring a paid marketer when you have zero revenue feels just as scary. And I&#x27;m not dancing on TikTok, that&#x27;s for sure.Have any of you actually taken on a marketing co-founder? What made you say yes to that person specifically? Was it their track record, the way they pitched, a trial period first?

Ask HN: How do you handle marketing as a solo technical founder?

Like most of you, I don&#x27;t use antivirus software and haven&#x27;t for years. Modern Macs (and PCs) are pretty secure by default.However, with the rise of AI-assisted exploits&#x2F;phishing and supply chain attacks, I&#x27;ve been reconsidering. We recently had an incident at work where CrowdStrike caught a RAT that a developer was inadvertently installing on their work computer.1. Would consumer antivirus &#x2F; EDR software even be good enough to block things like the Axios compromise?2. What do you recommend?

Ask HN: Has anyone reconsidered Antivirus software after recent security news?

Curious about projects where a volunteer doc writer could make some impact.

Any Open Source projects in need of documentation writer?

I&#x27;ve been testing Tor Browser on Android (rooted tablet + Bluetooth tether sniffer). Here&#x27;s what I found:These requests contain:· Your real IP address
· The .onion URL in the Referer header
· Tor Browser user-agentThe evidence (captured live):```
[18:12:10] 10.188.1.98 -&gt; 192.178.183.95 (Akamai)
[18:12:14] 10.188.1.98 -&gt; 142.251.14.95 (Google)
[18:12:22] 10.188.1.98 -&gt; 142.251.20.95 (Google)
```HTTP attempts: 5
HTTPS SNI captured: 0All plain text. No encryption. Tor not involved.What this means:Every time you use Tor Browser on Android, switch to Desktop Mode, and visit a .onion site, you&#x27;re broadcasting your real IP to Google, Amazon, and anyone monitoring your network.

Tor Browser on Android leaks IP in desktop mode

A database could be described crudely as a program for interfacing with data on disk. Most software is built to interface with databases although the software also has (often very capable) ways of interfacing with data on disk via filesystem interaction.My question is, what historically has driven the industry to focus on database-specific solutions, rather than on filesystem-specific solutions?This is not a rant against databases, but I do wonder why many major programming languages and frameworks (RoR Active Directory, C# EF, etc) have put much effort into making database interaction smooth from the perspective of the programming language, instead of putting effort into interacting with data on disk. Kind of an alternate reality sort of musing.

Ask HN: Why Databases Instead of Filesystem?

There is a product I really like. Love using it, love tinkering with it, and I buy into the overall vision. Founder has also always been pretty respectful publicly.But I recently came across some of their political views and now I just feel a bit weird about it. Nothing changed about the product, just how I feel.Do you separate the product from the founder?
Does this change whether you keep using something?
Or only if it starts showing up in the product?Don’t ask what the product is. It doesn’t matter and I’m not going to share.

Do founders' political views affect how you see a product?

For the longest time I thought I just hadn’t found the idea yet. I’d get excited about something, go a few layers deep, maybe even start planning it out… and then a week or two later I’d lose conviction. Then repeat the same thing with a different idea. It felt productive in the moment, but looking back it was just a loop.At some point I got tired of that and did something slightly different. Instead of picking one idea and obsessing over it, I started collecting them. Just dumping everything interesting I came across into one place, patterns, half-baked thoughts, things I’d normally ignore after a day. That eventually turned into a small database (I later called it StartupIdeasDB, mostly for myself), but the useful part wasn’t the tool,it was being able to see ideas next to each other without being emotionally attached to any one of them.And that’s where things started to shift. A lot of ideas that felt “solid” in isolation looked pretty weak when placed next to 20 similar ones. You start noticing how often the same concepts repeat with slightly different wording. You also start realising how easy it is to convince yourself something is good just because you spent a few hours thinking about it.The bigger realisation for me was that I was asking the wrong question the whole time. I kept asking “is this a good idea?” which is a very easy question to answer with bias. What actually helped more was asking “who is this immediately useful for, and how would it reach them?” Most ideas don’t survive that question.Also noticed that the difference between something sounding smart and something people will actually use is… huge. Way bigger than I assumed earlier.
Ironically, going through hundreds of ideas didn’t make me more creative or inspired. It made me a lot more skeptical. In a good way, I think.Now when I come across something interesting, my default instinct is to try and disprove it quickly instead of getting excited and building around it. Still figuring this out, but this shift alone has probably saved me months of going in the wrong direction.How are others here evaluate ideas before committing to them ?

I collected startup ideas. It changed how I think about ideas completely

There are essentially two established ways to use zooming in web interfaces today. They serve different purposes and make different tradeoffs. I built a third one, so I&#x27;ll try to be fair about what each does well and where it falls short.* Prezi
Prezi pioneered the zooming canvas for presentations and remains the market leader in that space. It recently added AI-powered generation and text editing tools. It&#x27;s a polished product with real traction.But Prezi is a closed platform, not a library. You can&#x27;t use its zoom engine in your own app. Pricing starts at $15&#x2F;month for meaningful features, and exporting to PowerPoint flattens all zoom effects into static slides. A recurring complaint from users is that the zooming and panning transitions cause motion sickness. And fundamentally, Prezi uses zoom as a storytelling device between pre-arranged frames. It&#x27;s not a navigation model. It&#x27;s a presentation model.* impress.js
impress.js brought Prezi-like zooming to the open web. It&#x27;s a presentation framework based on CSS3 transforms and transitions, directly inspired by Prezi. It was genuinely groundbreaking when it launched.
Its architecture is step-based: you position &quot;steps&quot; in 3D space and the camera moves between them. That&#x27;s great for presentations, but it doesn&#x27;t help you build an app where users navigate by zooming into content. impress.js has no concept of dynamically mounting views, managing zoom depth, or handling navigation state. It&#x27;s a slide deck engine with a zoom trick.* Zumly
This is what I built. Full disclosure: I&#x27;m the sole developer.
The idea is offering an alternative to traditional page navigation using zooming. You mark an element as zoomable, point it to a view, and Zumly handles the transition and inserts new views. That&#x27;s basically it.I started Zumly in 2020 after leaving behind Zircle UI (a Vue zooming library), trying to take what I learned further. Framework-agnostic, focused just on the zoom part. Since then I&#x27;ve rewritten the engine several times, changed the approach more than once. Only now I&#x27;m actually happy with how it feels.Views are dynamically mounted and unmounted during zoom transitions. In impress.js, all steps exist in the DOM simultaneously. In Zumly, you zoom into a trigger element, and the target view gets injected and scaled into place. This is closer to how routing works in SPAs than to how slide decks work.The landing page is built with Zumly itself so you can get the feel before touching any code.Curious if anyone else has thought about this space. What makes zooming UIs work or fail?Landing page (built with Zumly): https:&#x2F;&#x2F;zumerlab.github.io&#x2F;zumlyGitHub: https:&#x2F;&#x2F;github.com&#x2F;zumerlab&#x2F;zumly

Zooming UIs in 2026: Prezi, impress.js, and why I built something different

Something I’ve noticed (I’m using Claude subscription so no refunds but it applies to usage windows) is that sometimes AI makes mistakes so if something is important I tell Claude code to spin up a couple of sub agents and verify the information, often there will be a mistake and it gets rectified.It feels unfair I have to pay (or lose some usage) for this.Interested in other people’s thoughts.

Ask HN: Should AI credits be refunded on mistakes?

It seems that companies use VC funds as the main advertising for their company these days. In fact, without capital raising it is hard to set yourself apart from the rest of the budding companies started in their thousands each day.The receiving of VS funds now seems the only way one can create any virality in the marketplace. People still have a plea to the expert-from-out-of-town in that they feel... &quot;well VCs are very experienced people who know what to look for when investing&quot;, which is counter-intuitive when faced with the realisation that VCs fund 19 losers for every 1 unicorn.People want a company to be big before it has a marketplace-chance of being big. An press release of &quot;ABC Co., a startup tech firm, has a $25M investment from Vulture VC Fund resulting in a $1B valuation&quot; carries considerably more viral weight that a PR stating &quot;ABC Co., a startup tech firm, has a business plan that can change the world&quot;.And the VC industry is certainly promoting this line of thought... that the starting point of any future unicorn and VC funding go hand-in-hand. With the converse implications that if you have a great concept and execution path, that VC funds are a requirement.

Is VC the new PMF strategy?

TLDR: Forked pytorch and triton internals . Changed attention so its linear first layer , middle quadratic layer, last linear layer
Inference got much faster with a low perplexity hit in tests .Full attention O(n²): 17.96s &#x2F; 5.6 tok&#x2F;sHybridAttention O(n·W + n·D): 0.35s &#x2F; 286.6 tok&#x2F;sI have been building a small Rust focused language model from scratch in PyTorch. This is not a finetune. It is byte level, trained from random initialization on a Rust heavy corpus assembled here: https:&#x2F;&#x2F;codeberg.org&#x2F;JohannaJuntos&#x2F;SisyphusModel and training setupThe model has 25.6M parameters with a 512 context length. It uses a byte level vocabulary of 256, with 8 layers, 8 heads, and 512 dimensional embeddings. Positional embeddings are learned and the embedding and LM head weights are tied.Training ran for 30k steps on a 173.5M byte Rust corpus using a single RTX 4060 Ti 8GB.Final metrics were a train loss of 0.5834, validation loss of 0.8217, and perplexity of 2.15. The best validation loss occurred around step 18.5k, which suggests some late overfitting or plateau.ArchitectureThe model is a GPT style decoder, but replaces standard full attention with a HybridAttention block in each layer. This combines local windowed causal attention with a GRU like recurrent state path, along with a learned gate that mixes the two.The local path handles short range syntax, while the recurrent path carries compressed long range state. The gate bias is initialized to favor local attention early in training.Inference uses Triton kernels and custom torch.library ops.CorpusThe biggest gain came from corpus expansion.The run started with about 31MB from Rust official sources and major projects such as rustc, cargo, rust analyzer, tokio, serde, ripgrep, clap, and axum. The corpus was expanded to 173.5M bytes by cloning the top 500 crates, with 461 successful clones.This expansion had more impact than any architectural change.Inference performanceFull attention runs at about 5.6 tokens per second, while HybridAttention with KV cache reaches 286.6 tokens per second. This is about a 51x speedup with no visible quality loss.The KV cache uses a hot window of 64 tokens in VRAM, while older tokens are compressed to 8 bit magnitude and angle and can be selectively promoted back to full precision. This changes the effective complexity from quadratic to near linear for this setup.QualitySurface Rust syntax looks decent, and imports and function signatures are often plausible. Semantics are still weak, and repetition and recursive patterns are common. It looks like Rust, but does not reason well yet.What seems interestingThis project combines byte level Rust only pretraining from scratch, a hybrid local attention and recurrent architecture, large scale corpus expansion across the Rust ecosystem, and a practical KV cache paging strategy that delivers large speedups on consumer GPUs.Next stepsI plan to run ablations comparing hybrid attention against local only and recurrent only variants, evaluate checkpoints around 18.5k versus the final model, and add syntax level validation such as parsing and compiling generated code. I also want to explore scaling context length from 256 up to 2048 and test whether switching from byte level to BPE becomes worthwhile now that the corpus is larger.QuestionsFor small code models, which evaluations have been most useful beyond perplexity?Has anyone seen hybrid local plus recurrent attention work well for code generation?Given this setup, would you prioritize more tokens, longer context, or clean ablations first?

Hybrid Attention

Cuttle is a combat card game of mysterious origins dating back to the 1970&#x27;s. It feels like playing Magic or Yugioh!, but it&#x27;s played with a standard 52-card deck.https:&#x2F;&#x2F;cuttle.cards is an open source (code: https:&#x2F;&#x2F;github.com&#x2F;cuttle-cards&#x2F;cuttle) project that lets people play Cuttle online, for free.The project is under active development and has a competitive format with 4 seasonal championships each year, followed by a world championship tournament. The Cuttle World Championship is tomorrow, April 11th, at 1pm ET. You can watch it live on Twitch (https:&#x2F;&#x2F;twitch.tv&#x2F;cuttle_cards) or Youtube (https:&#x2F;&#x2F;www.youtube.com&#x2F;live&#x2F;mBj04alF1Z8?si=amIR5flmW7zewmL9) and chat with the players on the Cuttle discord (https:&#x2F;&#x2F;discord.gg&#x2F;46RTv7S4Hg).*RULES*
### GoalThe goal is to be the first player to have 21 or more points worth of point cards on your field. The first player to reach the goal wins immediately. One player (traditionally the dealer) is dealt 6 cards, and their opponent is dealt 5. The player with 5 cards goes first.### PlayOn your turn you must perform exactly one of the following actions:- *Draw:* Take one card from the Deck. You may not draw past the 8-card hand limit
- *Points:* Play a number card from your hand. Worth its rank in points, lasts until scrapped
- *Scuttle:* Scrap an opponent’s point card with a bigger one from your hand
 - If two number cards have the same rank, the suit breaks the tie
 - Suit order is (high) Spades ♠ &gt; Hearts ♥ &gt; Diamonds ♦&gt; Clubs ♣ (low)
- *Royal:* Play a face card for a persistent benefit based on rank (lasts until scrapped)
- *Glasses:* Play an Eight to reveal your opponent’s hand (lasts until scrapped)
- *One-Off:* Scrap a number card for an effect based on the rank of the card.## RoyalsRoyals (Kings, Queens, and Jacks) may be played to the field for a persistent benefit that remains in effect until the card is scrapped. Each Royal gives a different effect.*King:* Reduce the number of points you need to win (21, 14, 10, 5, 0 points with 0, 1, 2, 3, 4 kings)*Queen:* Protects your other cards from being targeted by the effects of other cards. This protects your cards against 2’s (both effects), 9’s, and Jacks, but not scuttling.*Jack:* Play on an opponent’s point card to steal it. Point card returns to opponent if the jack is scrapped or if another jack is used to steal it back.### One-OffsNumber cards (except 8’s and 10’s) can be played for a One-Off effect, which scraps the card for an effect based on the rank of the card played. Whenever a one-off is played, the other player may counter it using a two to cancel the effect.- *Ace:* Scrap all point cards on the field
- *Two:* Twos have two alternative one-off effects:
 - Counter target One-Off Effect (Played immediately in response to a one-off)
 - Scrap target Royal
- *Three:* Choose a card in the scrap pile and put it in your hand
- *Four:* Your opponent discards two cards of their choice
- *Five:* Discard one card and then draw three cards
- *Six:* Scrap all Royals and Glasses Eights on the field
- *Seven:* Choose one of the top two cards from the deck and play it however you choose.
- *Nine:* Return a card from your opponent’s field to their hand. They cannot play it next turn.

Open Source card game cuttle.cards has its world championship Saturday at 1pm ET

Ofc after you die, doesnt matter in the first place, but to be honest, I&#x27;ve been thinking about what my wife would have to deal with digitally when I&#x27;m gone. All accounts on the internet, my old pictures in Google Drive, documentation and things like that.I have password manager, so I pass this to my wife it can be easy to reach the profiles, but what else can I do to help &#x2F; easier the whole process?

Ask HN: How do you manage your digital legacy for after you die?

I am looking for local-first a meetings recorder that- supports on-device models- transcribes to MD- detects ongoing meetings- open sourceI was actively using Char&#x2F;Hyprnote, but they have stopped supporting local models. Big fan of handy.computer UX, maybe something along those lines.Anyone using something good?