The Waymo World Model

Hello everyone!I&#x27;m a solo developer who&#x27;s been doing UI&#x2F;UX work since 2007. Over the years, I watched design tools evolve from lightweight products into bloated feature-heavy platforms. I kept finding myself using a small amount of the features while the rest just mostly got in the way.So a few years ago I set out to build a design tool just like I wanted. So I built Vecti with what I actually need: pixel-perfect grid snapping, a performant canvas renderer, shared asset libraries, and export&#x2F;presentation features. No collaborative whiteboarding. No plugin ecosystem. No enterprise features. Just the design loop.Four years later, I can proudly show it off. Built and hosted in the EU with European privacy regulations. Free tier available (no credit card, one editor forever).On privacy: I use some basic analytics (page views, referrers) but zero tracking inside the app itself. No session recordings, no behavior analytics, no third-party scripts beyond the essentials.If you&#x27;re a solo designer or small team who wants a tool that stays out of your way, I&#x27;d genuinely appreciate your feedback:
<a href="https:&#x2F;&#x2F;vecti.com" rel="nofollow">https:&#x2F;&#x2F;vecti.com</a>Happy to answer questions about the tech stack, architecture decisions, why certain features didn&#x27;t make the cut, or what&#x27;s next.

Show HN: I spent 4 years building a UI design tool with only the features I use

How to effectively write quality code with AI

Microsoft open-sources LiteBox, a security-focused library OS

Sheldon Brown's Bicycle Technical Info

Learning from context is harder than we thought

I spent 5 years in DevOps – Solutions engineering gave me what I was missing

Due to bike-induced concussions, I&#x27;ve been worried for a while about losing my memory and not being able to log back in.I combined shamir secret sharing (hashicorp vault&#x27;s implementation) with age-encryption, and packaged it using WASM for a neat in-browser offline UX.The idea is that if something happens to me, my friends and family would help me get back access to the data that matters most to me. 5 out of 7 friends need to agree for the vault to unlock.Try out the demo in the website, it runs entirely in your browser!

Show HN: If you lose your memory, how to regain access to your computer?

Understanding Neural Network, Visually

I now assume that all ads on Apple news are scams

An Update on Heroku

Hi HN! Smooth CLI (<a href="https:&#x2F;&#x2F;www.smooth.sh">https:&#x2F;&#x2F;www.smooth.sh</a>) is a browser that agents like Claude Code can use to navigate the web reliably, quickly, and affordably. It lets agents specify tasks using natural language, hiding UI complexity, and allowing them to focus on higher-level intents to carry out complex web tasks. It can also use your IP address while running browsers in the cloud, which helps a lot with roadblocks like captchas (<a href="https:&#x2F;&#x2F;docs.smooth.sh&#x2F;features&#x2F;use-my-ip">https:&#x2F;&#x2F;docs.smooth.sh&#x2F;features&#x2F;use-my-ip</a>).Here’s a demo: <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=62jthcU705k" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=62jthcU705k</a> Docs start at <a href="https:&#x2F;&#x2F;docs.smooth.sh">https:&#x2F;&#x2F;docs.smooth.sh</a>.Agents like Claude Code, etc are amazing but mostly restrained to the CLI, while a ton of valuable work needs a browser. This is a fundamental limitation to what these agents can do.So far, attempts to add browsers to these agents (Claude’s built-in --chrome, Playwright MCP, agent-browser, etc.) all have interfaces that are unnatural for browsing. They expose hundreds of tools - e.g. click, type, select, etc - and the action space is too complex. (For an example, see the low-level details listed at <a href="https:&#x2F;&#x2F;github.com&#x2F;vercel-labs&#x2F;agent-browser" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;vercel-labs&#x2F;agent-browser</a>). Also, they don’t handle the billion edge cases of the internet like iframes nested in iframes nested in shadow-doms and so on. The internet is super messy! Tools that rely on the accessibility tree, in particular, unfortunately do not work for a lot of websites.We believe that these tools are at the wrong level of abstraction: they make the agent focus on UI details instead of the task to be accomplished.Using a giant general-purpose model like Opus to click on buttons and fill out forms ends up being slow and expensive. The context window gets bogged down with details like clicks and keystrokes, and the model has to figure out how to do browser navigation each time. A smaller model in a system specifically designed for browsing can actually do this much better and at a fraction of the cost and latency.Security matters too - probably more than people realize. When you run an agent on the web, you should treat it like an untrusted actor. It should access the web using a sandboxed machine and have minimal permissions by default. Virtual browsers are the perfect environment for that. There’s a good write up by Paul Kinlan that explains this very well (see <a href="https:&#x2F;&#x2F;aifoc.us&#x2F;the-browser-is-the-sandbox" rel="nofollow">https:&#x2F;&#x2F;aifoc.us&#x2F;the-browser-is-the-sandbox</a> and <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46762150">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46762150</a>). Browsers were built to interact with untrusted software safely. They’re an isolation boundary that already works.Smooth CLI is a browser designed for agents based on what they’re good at. We expose a higher-level interface to let the agent think in terms of goals and tasks, not low-level details.For example, instead of this:<pre><code> click(x=342, y=128)
 type(&quot;search query&quot;)
 click(x=401, y=130)
 scroll(down=500)
 click(x=220, y=340)
 ...50 more steps
</code></pre>
Your agent just says:<pre><code> Search for flights from NYC to LA and find the cheapest option
</code></pre>
Agents like Claude Code can use the Smooth CLI to extract hard-to-reach data, fill-in forms, download files, interact with dynamic content, handle authentication, vibe-test apps, and a lot more.Smooth enables agents to launch as many browsers and tasks as they want, autonomously, and on-demand. If the agent is carrying out work on someone’s behalf, the agent’s browser presents itself to the web as a device on the user’s network. The need for this feature may diminish over time, but for now it’s a necessary primitive. To support this, Smooth offers a “self” proxy that creates a secure tunnel and routes all browser traffic through your machine’s IP address (<a href="https:&#x2F;&#x2F;docs.smooth.sh&#x2F;features&#x2F;use-my-ip">https:&#x2F;&#x2F;docs.smooth.sh&#x2F;features&#x2F;use-my-ip</a>). This is one of our favorite features because it makes the agent look like it’s running on your machine, while keeping all the benefits of running in the cloud.We also take away as much security responsibility from the agent as possible. The agent should not be aware of authentication details or be responsible for handling malicious behavior such as prompt injections. While some security responsibility will always remain with the agent, the browser should minimize this burden as much as possible.We’re biased of course, but in our tests, running Claude with Smooth CLI has been 20x faster and 5x cheaper than Claude Code with the --chrome flag (<a href="https:&#x2F;&#x2F;www.smooth.sh&#x2F;images&#x2F;comparison.gif">https:&#x2F;&#x2F;www.smooth.sh&#x2F;images&#x2F;comparison.gif</a>). Happy to explain further how we’ve tested this and to answer any questions about it!Instructions to install: <a href="https:&#x2F;&#x2F;docs.smooth.sh&#x2F;cli">https:&#x2F;&#x2F;docs.smooth.sh&#x2F;cli</a>. Plans and pricing: <a href="https:&#x2F;&#x2F;docs.smooth.sh&#x2F;pricing">https:&#x2F;&#x2F;docs.smooth.sh&#x2F;pricing</a>.It’s free to try, and we&#x27;d love to get feedback&#x2F;ideas if you give it a go :)We’d love to hear what you think, especially if you’ve tried using browsers with AI agents. Happy to answer questions, dig into tradeoffs, or explain any part of the design and implementation!

Show HN: Smooth CLI – Token-efficient browser for AI agents

Hi HN,My friend and I have been experimenting with using LLMs to reason about biotech stocks. Unlike many other sectors, Biotech trading is largely event-driven: FDA decisions, clinical trial readouts, safety updates, or changes in trial design can cause a stock to 3x in a single day (<a href="https:&#x2F;&#x2F;www.biotradingarena.com&#x2F;cases&#x2F;MDGL_2023-12-14_Resmetirom_NASH" rel="nofollow">https:&#x2F;&#x2F;www.biotradingarena.com&#x2F;cases&#x2F;MDGL_2023-12-14_Resmet...</a>).Interpreting these ‘catalysts,’ which comes in the form of a press release, usually requires analysts with previous expertise in biology or medicine. A catalyst that sounds “positive” can still lead to a selloff if, for example:
the effect size is weaker than expected- results apply only to a narrow subgroup- endpoints don’t meaningfully de-risk later phases,- the readout doesn’t materially change approval odds.To explore this, we built BioTradingArena, a benchmark for evaluating how well LLMs can interpret biotech catalysts and predict stock reactions. Given only the catalyst and the information available before the date of the press release (trial design, prior data, PubMed articles, and market expectations), the benchmark tests to see how accurate the model is at predicting the stock movement for when the catalyst is released.The benchmark currently includes 317 historical catalysts. We also created subsets for specific indications (with the largest in Oncology) as different indications often have different patterns. We plan to add more catalysts to the public dataset over the next few weeks. The dataset spans companies of different sizes and creates an adjusted score, since large-cap biotech tends to exhibit much lower volatility than small and mid-cap names.Each row of data includes:- Real historical biotech catalysts (Phase 1–3 readouts, FDA actions, etc.) and pricing data from the day before, and the day of the catalyst- Linked Clinical Trial data, and PubMed pdfsNote, there are may exist some fairly obvious problems with our approach. First, many clinical trial press releases are likely already included in the LLMs’ pretraining data. While we try to reduce this by ‘de-identifying each press release’, and providing only the data available to the LLM up to the date of the catalyst, there are obviously some uncertainties about whether this is sufficient.We’ve been using this benchmark to test prompting strategies and model families. Results so far are mixed but interesting as the most reliable approach we found was to use LLMs to quantify qualitative features and then a linear regression of these features, rather than direct price prediction.Just wanted to share this with HN. I built a playground link for those of you who would like to play around with it in a sandbox. Would love to hear some ideas and hope people can play around with this!