PH热榜 | 2026-06-12

@derickd Man it's crazy how cheap this is. Can i get a mail where i can contact you can talk more about the business side of things? Would love to talk.

@derickd Congratulations on the launch, Derick! 🎉

The pricing is genuinely eye catching, and I like the API first approach. Making e signatures a built in feature instead of a cost center makes a lot of sense for growing SaaS products.

One question: among your early customers, what has been the biggest reason for switching from existing providers? Was it mainly cost, developer experience, implementation speed, or something else entirely?

Wishing you and the team a fantastic launch!

@derickd Congrats, Derick! 3¢ an envelope is an absolute game-changer for startups escaping DocuSign's insane pricing.

I would like to join you on fast scale, as the big hurdle to this point is to offer massive outbound targeting to HR, fintech, and proptech startups who hate high API costs. I can manually source 200 hyper-targeted developer leads and handle all cold outreach replies for you. Let me handle growth while you code!
I look forward to hearing from you. Thanks

@ankit_narang1 GREAT question yes obviuosly cost is very compelling, but outside that, we find our top customer needs are, in this order:

1. Legality in your country .. this is binary, you either are or you arent. E.g. we're not legal in China, but we are in 55+ other countries including EU, LATAM, US, and a lot of APAC.

2. API capability (usually this translates to white label) .. this is an area where we really shine. We built what we always wanted and I think it shows.. we're proud of it.

@artstavenka1 Hi Art. Yup, you sure can. One of our key differentiors at Firma.dev is that we have workspaces (like, at all). Docusign has the capability in its highest tier enterprise plan$.

We created Firma.dev and workspaces because at our previous startup, we wanted to offer e-sign for free to our customers (we never could before), and do it in a secure way.

So the guardrail for the MCP ... itll ask you what workspace you want to engage with, and you have to "manually" switch workspaces. So if you're working with a particular set of customers, one customer, or a team, it works. Its like this: Use workspace Sales Team and send out the Customer Agreement to Derick at support@firma.dev

... make sense?

We're seeing a growing amount of our customers (especially on the real SMB side) using agentic workflows like that.

@margarita_s88 Thanks Margarita! We started doing customer testimonials and mini case studies and prioritizing "realness" over quality lol so they look a little homemade but at least you can tell they are real customers! Good luck on the launch Alconost looks useful Im passing this to our team as well.

@nimmy_fanimi Hi Nimmy! Thanks. Documentation is our core product.. since all we make is an API, we have to stay on top of that. We have a stand up meeting every single morning about how to better improve our API docs ... one of the questions we ask each other is "whats 1 thing you can fix today, even if its small, you have to improve something". So that's it .. its like exercise .. just gotta do it every day, even if you don't wanna.

@thamibenjelloun Hi Thami. Yes, our API is very powerful, and one of the largest on the market. We have everything you asked about: multiple signers with order, reminders (from us or you), custom domains, graphics, etc per workspace, and tons of webhooks (and those are even per workspace). Thanks for taking a look!

@busra_seker1 Yah thanks Büşra .. dont get me started on what real volume means to Docusign etc. Its crazy. Even then, they don't come down to our prices. Thanks for taking a look. 😁

@borrellbr Hi Ignacio! Thanks for your comment.

Pretty evenly split among 3 groups.

• SMB - midcap that are creating mini apps for their own internal workflows. Mostly migrating from purely manual workflows b/c other solutions are too expensive.

• Established/VC backed SaaSs - Migrating from Docusign, DocuXXXX, (all the docu's) because we offer a fully white labeled product finally thats doesn't kill your budget

• Startups - People building little apps that need a quick cheap way for their users to sign docs and theres no way that they're gonna get on the phone with Docusign.

@farrukh_butt1 Thanks Farrukh. We're devs, and had the pain point first hand of having TERRIBLE APIs with the existing e-sign providers, so we built it how we would want. The thing about e-sign is, there's a lot of little laws you need to follow and be aware of .. audit trails are part of that. We have fully legal, compliant audit trails (just like Docusign) in 55+ countries. Okay fine if you need documents in China, use Docusign and pay $5 an envelope. But ours are totallly compliant and you can even configure them via the API / GUI.

@darly_selby YES DARLY THANK YOU--I feel seen 😅 that's one of our true core features is that the entire API(branding, webhooks, keys, everything) can be silo'd. E.g. You're a multi tenant saas and a new customer signs up, you auto-provision a new workspace, they can create their own templates, domains, etc, but yah you get it.

But to answer your question .. yah it was built from our own (painful) experiences on the customer side.

It came from that feature not existing in the open market when we were building an HR app. At that company, were using a Docusign knockoff and their API cost around 50 cents per doc, but we had to go an manually (through the GUI) make a team, and then our support staff would have to manually create their templates for them .. super labor intensive.

We didn't even know what to call them b/c the concept didn't exist, when all those other e-sign companies are built around the concept of a GUI and individual users sending docs .. not an API.

Thanks for taking a look.

@ada_johnsen Yup! And you can get started super easy .. versus talking to a sales team and getting quotes for API usage. Thanks for looking! 😁

@natalia_iankovych Hi Natalia! Great question. We probably aren't the best fit for you, unless you're retooling some of your workflows, e.g. building internal mini apps, or have software using an API. We do have a full powered GUI (that we're proud of) .. but white labeling via API for super cheap volume senders will always be our focus. 98% of our users send via API, 2% GUI.

We also play really well with things like N8N, Salesforce, etc. Our "self service" option is mostly used by people that connect our MCP server right into their Claude/ChatGPT .. but honestly thats pretty case dependent, you know?

However, if you find yourself thinking about processes in your org, and e-sign is part of that, hit us up.

@theomkarbirje Great idea, thanks for building this. I definitely want to try it.

The biggest pain for me right now is that I usually have to do this through Inspect, and it takes too much time. Styles are not included, so I often end up combining a screenshot with copied source code just to explain one UI change to an agent.

This is exactly the kind of friction that should disappear.

@theomkarbirje Love this. The fact that it came from a real workflow pain makes it immediately click.

I've lost count of how many times I've had to fight an AI agent just to modify a specific UI element, and the back-and-forth can get ridiculous. Giving agents structured context instead of making them guess from screenshots feels like a very obvious solution in hindsight.

Congrats on the launch. Curious to learn more about the business behind it as well. What's the best email to reach you on? Would love to connect and explore a few ideas.

@theomkarbirje Congratulations on the launch! 🎉

This solves a very real problem. I have definitely spent more time than I'd like trying to explain to an AI agent which button, section, or component I wanted changed.

One question: have you found that providing structured element context significantly improves the accuracy of AI-generated UI changes compared to screenshots alone? If so, what kind of reduction in prompts or iterations are users typically seeing?

Wishing you a fantastic launch and plenty of happy users! 🚀

@jaime_iniesta Tidewave looks great when its pointer works with your stack, and Qursor is aiming to be the “works everywhere” version so you can point at any UI in Chrome and still grab clean selectors and styles for your agent, even on projects outside those supported frameworks.

@chintant As much as I know, Webflow is a visual editor where you design directly in the dashboard, not an AI agent. Qursor is most helpful when you are working with AI coding agents or chat-based tools and need to send them precise UI context.

@saad_nadeem3 Qursor does not care which framework you use because it reads from the rendered DOM, so when you point at something, it grabs the real element with its selectors and styles.

@alpertayfurr Thanks a lot for saying it so clearly. That is exactly the gap I am trying to close

@oktay_atalay_ Love this, thanks a lot for taking the time to write it. That “no, the button next to that one” loop is exactly what I was stuck in, so it is really cool to hear that the structured context idea resonates.

@dren0r Totally, MCP DevTools is great when your agent already has full page access wired up. Qursor is more for cases where you are just chatting with an AI coding agent or using different tools, and you want a fast, framework‑agnostic way to point at a specific element and hand over clean selectors, styles or HTML without wiring anything.

@henry_habib Qursor works the same way on internal tools and SaaS dashboards as long as you can open them in Chrome, so you can point at elements in your admin panels or internal UIs just like on public sites.

@luca_capone You are describing exactly the same leak I ran into. With Qursor, it is not just the visual; it grabs the element plus structured context around it: selectors, classes, inline styles, and layout info, so the agent can see both what it looks like and where it lives in the DOM.

In short, the thing you do in ClaudeCode of describing, instead you start Qursor and point to which element you wanted to change, click and annotate here. You can do multiple annotations on this page and then copy all annotations for Claude Code to do.

@alon4 Love hearing this, that is exactly the loop that pushed me to build Qursor.

@ianhxu Great question. In practice the copied context is surprisingly small – usually just a few hundred tokens even on a busy page – so it is tiny compared to a screenshot and nowhere near half of an agent’s window. Also, it has 4 modes, so you can go even smaller or increase if you wanted to give clearer context.

@farrukh_butt1 Client feedback is where this hurts a lot. With Qursor, they can still click on “change this section,” but instead of a blurry screenshot, you get the exact element with selectors and styles you can hand straight to a dev or an agent.

@thinksdylan The framing here makes sense as a response to the compression of startup building cycles, but the real long-term challenge will likely be market integrity at scale.

Once fundraising, GTM execution, and bounties sit in one ecosystem, the definition of “signal” becomes the product itself.

Curious how you’re thinking about separating durable traction signals (retention, revenue quality) from optimized short-term performance metrics as participation grows globally.

@thinksdylan Thank you for creating Pond. In all honesty, traditional fundraising platforms have not kept up with the rapid changes AI has introduced to the startup environment. Great ideas fall by the wayside because of commercialisation challenges that founders find insurmountable. Thank you.

@thinksdylan congrats on the launch! will you guys be operating in the US?

@thamibenjelloun Great question.

For founders, verification starts with real business traction. Today, we focus on signals like revenue, MRR, user growth, and connected data sources such as Stripe or GA where available.

The goal is to help investors look beyond pitch decks and evaluate startups based on actual performance. We’re also working toward making these metrics refresh regularly so founder profiles stay up to date as the company grows.

Feel free to check out Pond and send over any feedback.

@mert__33 Totally agree — getting the first customers is often harder than building the product.

That’s why we see fundraising and GTM as connected. Pond Markets helps founders turn traction into capital, while Pond Growth and Bounties help with feedback, early users, campaigns, and distribution support.

We want to help startups keep growing after they raise, not just complete a round.

@dhiraj_patel5 We’ve recently supported bounties that helped companies close deals more efficiently and create a stronger sense of urgency among prospective customers.

For example, one startup posted a bounty with us and delegated the community to conduct competitive analysis and research on their target accounts. The goal was to identify whether those prospects were already using services similar to the startup’s offering.

The startup also asked community members to share their findings publicly on social media. For instance, if the community discovered that a prospect’s competitors were already using a particular service, they would post about it and tag the prospect. A typical post might say:

“Based on our recent research, we found that [Company A] is using this service, and it has helped them improve key metrics by X%.”

The startup then used these research findings to strengthen its sales outreach. Their sales team could approach prospects with a message like:

“Someone recently highlighted that your competitor is using a solution similar to ours, and reports suggest it has improved their efficiency by X%. Our platform delivers even better results. Would you be interested in exploring how we can help you outperform your competitors in this area?”

@artem_myronchuk Exactly. If a startup has real traction, revenue, and user growth, it should be able to access capital and support regardless of where the founder is based. That’s a big part of why we built Pond.

@marianna_tymchuk Thank you, Marianna — that’s exactly the problem we’re trying to solve.

Founders today don’t just need capital. They also need customers, contributors, feedback, and execution support. Pond brings those pieces together so startups can move from traction to scale much faster.

@busmark_w_nika Hello there Nika. Great question! Quick reframe first, because it's kind of the whole point .

HyperSleep doesn't just block during set hours. It keeps your apps locked until it verifies you've actually slept (Google Sleep API + motion + light + usage signals). So even at 2am, opening Instagram won't work until you've hit your sleep goal, there's no timer to wait out.

Beyond that core:

  • 🌙 Auto-starts at your bedtime. No remembering to turn it on

  • 🎯 You choose which apps to lock and your own sleep goal

  • 🔥 Streak tracking + a morning recap - sleep time, quality %, and how you did vs your goal

  • 🆓 Real-life overrides. A 5-min "quick break" (adds to your goal) or a full override with an accountability cost, so it's flexible, not a prison

  •🔐 100% on-device; no account, nothing leaves your phone

Love what you're building with your minimalist phone app. We're chasing the same goal from different angles

@hypersleep Congrats on the launch, Roddy! Flipping the incentive structure to outcome-based rewards is a super interesting approach. I actually just ran the screen-time blocker and habit-tracking niche through my tool, MarketGapAI, which analyzes thousands of complaints across Reddit and the Google Play Store. Interestingly, the number one reason users delete or give up on strict blocking apps is false positives - like the app failing to unlock because a sensor misread passive awake time (like reading a physical book) as sleep, creating massive user frustration.

Curious to know - how does HyperSleep handle manual overrides or edge cases where the 70% confidence threshold misses the mark, or are you focusing on dialing in the sensor calibration first?

@kate_ramakaieva Thank you Kate🙏 Not yet. And that's by design for v1: HyperSleep runs entirely on your phone's own sensors (Sleep API + motion + light + usage), so it works for everyone with zero extra hardware. No watch to buy or keep charged.

That said, native Wear OS / smartwatch support is firmly on the roadmap — a watch's heart-rate and movement data would make the sleep verification even tighter, so it's a natural next layer.

Out of curiosity, what watch are you on? Helps me figure out which to prioritize 🙂

@harini_mukesh Haha, the fact that you're discovering a sleep app while scrolling Product Hunt past bedtime might be the most on-brand thing all launch 😅 You're exactly who I built this for.

Thank you for the kind words 🙏 iOS isn't here yet (Android first for now), but it's firmly on the roadmap and I'd love to have you try it. Want me to ping you the second it lands?

Also, a big update is dropping very soon with more features on the way, so it only gets better from here. Would love your honest take once you're in.

@wasil_abdal Thank you Wasil🙏. Frankly, the most important question here, so I'm really glad you asked.

Short version: you're never trapped. It's friction, not a prison.

🚨 Only the apps you pick are ever locked; calls, alarms, messages, maps, all of it stays open. A 3am emergency is never an issue.

  🆓 For the blocked apps, there's an "I need my phone" override right on the lock screen, a 5-min quick break or a full override that ends the session. Your call, anytime.

 🌙 Overrides carry a small accountability cost (a quick break nudges your goal) so cheating isn't frictionless but you're always the one in control.

And you're spot on that sensors aren't perfect. Insomnia and travel are exactly the cases no detector nails. That's precisely why the override exists. I'd rather be honest (~70% confidence + an easy escape hatch) than pretend it's a flawless sleep lab.

Appreciate the privacy callout too: local-first was a hill I'd die on 🙂 Curious though: does that balance feel right to you, or would you want the escape hatch even more frictionless? Still tuning it.

@alpertayfurr Exactly this. You put it better than I usually do 🙏

"Be more productive" is too abstract to lose to at 1am. Sleep is concrete: you either hit your hours or you didn't, no negotiating with yourself. And it's a goal you already want; the app isn't imposing some new discipline, it's enforcing the one you'd choose anyway when you're rested enough to think straight.

 It's also why I dodged the guilt angle most blockers lean on. Not "you're weak, scroll less". It's "sleep, and your apps are right there waiting." A reward you walk toward, not a punishment you endure.

Genuine question: do you think the boundary would feel just as meaningful tied to other behaviors someday (a workout, a morning routine), or does sleep hit different because it's the one we all quietly fail at?

@gabriella_anjani This honestly made my day 🙏 You just described exactly why I built it; I was the guy who'd set a "limit," tap "ignore," and resurface from Instagram at 2am wondering where the night went. The timer was never the problem. I was 😅.  So, I built something I couldn't argue with.

And local-first was non-negotiable: your sleep patterns are nobody's business but yours. No account, nothing leaves the phone.

Genuine question, since you resonate with it, what's your 1am app of choice? The one you know you should put down but somehow don't 😅

@oshylabs Thanks Arnold. And coming from someone in the recovery space, that means a lot. You're seeing the same thing that motivated this: short sleep wrecks the next day, and willpower-based blockers fail exactly when you're most tired. The whole bet is that earning access beats restricting it.

On shift workers and split sleepers; great question, because you're right that the "one clean 11pm–7am block" assumption breaks for a lot of people. Two things help here:

1. It's not tied to the clock. You can start a session manually whenever your sleep window is; 9am after a night shift works exactly the same as 11pm. The schedule is yours to set, not a fixed "night."

2. The goal is total verified sleep, not one unbroken block. We confirm sleep in bouts (20+ continuous min of fused signals to filter out noise), and what matters is accumulating toward your goal, so a split night reads as sleep, not a fail.

That said, you've hit on something we want to go deeper on: biphasic/segmented sleep and rotating shift patterns are their own design problem, and "what counts as one night" gets genuinely interesting there. Would love to pick your brain on what good recovery-aware handling looks like for the lifters you work with.

Appreciate the thoughtful launch support which is exactly the kind of feedback that sharpens the product. 🙏

@louislecat Great question Louis. Nope, you just keep your phone on your nightstand like usual or bedside table, wherever. No need to put it on the bed or under your pillow. HyperSleep reads motion, light, and Google's Sleep API in the background to confirm you slept, then unlocks your apps in the morning. No wearable, nothing on the bed. 😴

@umermedia Hey Umer! Fair question. There are a lot of screen-time apps out there 🙂

Here's the core difference:

Most blockers are time-based. HyperSleep is outcome-based.

Typical apps lock your apps on a schedule or a timer but a timer doesn't care whether you actually slept, and you can usually just tap "ignore" or "5 more minutes" until the willpower runs out.

HyperSleep gates access on verified sleep, not the clock. HyperSleep doesn't need willpower, it uses your body. We use multi-sensor detection (Google's Sleep API + motion + light + usage) to confirm you actually slept before your apps unlock. You can't snooze your way past it.

The psychology flip is the real magic: instead of restricting you ("you've used your 30 min, no more"), you earn your scroll time by sleeping. It feels like a reward, not a punishment, which is the difference between an app you delete in a week and one that actually changes the habit.

And for the inevitable "but what if I really need my phone?"; there are built-in overrides, so it nudges rather than imprisons.

What apps have you tried? Genuinely curious what made you bounce off them.

@farrukh_butt1 Hi! Great instinct. And good news, transparency is already built in, both live and in the morning.

While a session is active, open your phone anytime and you'll see the Active Sleep Ring - your live sleep progress. The moment the phone detects motion, the timer pauses and switches to "Detecting sleep…" until you've settled and set the phone aside. So you're never guessing what HyperSleep thinks you're doing; it's showing you, in real time.

In the morning, the unlock screen gives you the full summary: total sleep (e.g. 7h 12m), a quality %, how you did vs your goal, and your streak. So it's never a silent unlock. You always see the result you earned.

And we're making a good thing even better: a coming update adds an even more detailed breakdown - a signal-by-signal view (Sleep API ✓ / motion / light / phone-untouched) with the confidence score behind each unlock, so you can see exactly which signals agreed.

Appreciate you raising it. Building trust through transparency is exactly the bar we hold ourselves to. 🙌

@tarunmangukiya Congrats on the launch, Tarun! The interface looks incredibly clean. I actually just ran the cross-border platform payout niche through my tool, MarketGapAI, which scrapes complaints across Upwork and Trustpilot. Interestingly, a massive pain point for creators and freelancers right now is unexpectedly high hidden conversion fees and sudden account freezes when trying to move USD to local accounts.

Curious to know - how does KOSH protect users from those sudden compliance pauses, or are you primarily focusing on the advertising credit card side of the equation for now?

@tarunmangukiya Congrats on the launch, Tarun.

This is a problem I've seen come up repeatedly for founders, freelancers, and online businesses across Asia. Getting paid internationally sounds simple until you actually start dealing with banking restrictions, transfers, and platform-specific headaches.

I particularly like that you've combined receiving, holding, and spending into a single workflow rather than treating them as separate products.

Would love to learn more about how you're approaching the business and growth side of KOSH. What's the best email to reach you on? I think it would be interesting to connect and exchange ideas.

@tarunmangukiya Congratulations for the launch, Tarun.

@sahil_sen Thank you so much. How's that KOSH card looking?

@mahdi_nouri We don't have cashback currently. But we do have referrral rewards. :) You get $5 for each successful referrals.

@soediono_lenz Thank you so much. It's been really great working with you during beta.

@arcinston How is the app so far?

@varun_trivedi2 thank you so much broo!! you've been our early user. really appreciate all the feedbacks.

@bobsworkshop  @kemone_phillips Congrats on launching! How does the 'Behavioral DNA profiling' learn my habits—does it analyze git history locally, or does it learn iteratively from active session approvals?

@anand_thakkar1 Great breakdown of the tradeoffs you get the space. To answer your question: it's all three, but layered.

Project context uses RAG over the codebase that handles the "what are you building" awareness. Combined with summarization and dynamic file mapping. But the personalization layer is a completely separate proprietary architecture. Our Frank Engine (built alongside NVIDIA and Google Cloud) maintains a dual-structure memory system short-term interaction patterns and long-term behavioral DNA that profiles how you think, how you make decisions, and what approaches have failed for you before. It's not retrieval. It's not fine-tuning. It's a purpose-built personality engine that self learns your engineering philosophy over time so Bob's responses aren't just contextually accurate they're you-accurate.

@keirodev Really appreciate you auditing the repo that's exactly the kind of scrutiny we welcome. To answer directly: you choose your storage tier. Tier 1 is fully local your conversations, behavioral DNA, project index, everything lives in ~/.bob/ on your machine. Firebase doesn't exist in that world. Tier 3 is opt-in (NON DEFAULT) sync across our ecosystem (web, mobile, CLI) for users who want cross-device continuity and even then, your source code is never transmitted. Only conversation context and profile data sync. The boundary is: your decision to to activate the explicit gate between "nothing leaves" and "I choose to sync." No telemetry, no silent uploads, no gray areas.

@henry_habib Hey Henry! Thanks for the kind words 🙏 Entry level (8GB RAM, integrated GPU): MacBook Neo, MacBook Air, or any budget laptop runs Gemma 4 E4B or Mistral 7B. Responses in 2–5 seconds. Totally usable for daily coding. Sweet spot (16–32GB RAM, RTX 5060–5070 Ti): This is where most developers land. Runs Gemma 4 12B or DeepSeek-R1 7B. Responses in 1–3 seconds. Feels like pair programming in real-time. Powerhouse (32–64GB RAM, RTX 5080–5090): Runs 70B parameter models. Sub-2-second responses. Can also serve your entire team via SovereignLink...one machine, multiple developers connecting remotely. Happy to help you figure out the ideal setup for your specific hardware if you want to drop your specs!

@ianhxu We've tested it on the NVIDIA RTX 5070 Ti (via Acer Predator) and the NVIDIA RTX 4050 (via HP Victus). On the RTX 5070 Ti, responses come back in 2-4 seconds with a 7B model — fast enough to stay in your flow while coding. The 4050 is slightly slower (4-8 seconds) but still very usable for iterative chat. For Apple Silicon machines like the MacBook Neo (A18 Pro, 8GB unified), a model like Gemma 4 E4B fits comfortably in memory. Based on the hardware specs and Apple's Neural Engine acceleration, we'd expect similar response times to the 4050 solidly in the keep in the loop while coding range.

@alina_tyslenok_ Thank you! That's exactly the problem we built for most of us aren't coding 8 hours straight, we're grabbing 45 minutes between meetings or an hour after the kids are asleep. Bob remembers everything between sessions so you never waste those precious minutes re-explaining context. Pick up exactly where you left off, from any device.

@jared_salois Great question. The CLI is distributed via npm, so users control when they update...pnpm update -g @bobsworkshop/cli when they're ready. We version intentionally: core features are stable across versions, and new capabilities are additive, not breaking. If you're on v0.5.6 and we're demoing v0.7.1, your existing commands still work identically you just don't have the new ones yet. We also flag significant updates in the terminal on launch so users know when something meaningful is available without forcing anything. Your machine, your pace.

@furkan_topcuoglu It really is a fundamentally different tradeoff. Cloud agents charge you more the better you get at using them we think that's backwards. With local-first, your costs stay flat (zero), your latency drops, and your data never becomes someone else's training set. All upside, no compromise.

@busra_seker1 You nailed it that hesitation is real and it's justified. Your codebase is your IP. With Bob, there's nothing to "give access" to because the AI runs on your machine alongside your files. No API call carrying your source code across the internet. The trust problem disappears when the architecture makes exposure physically impossible.

@amanda_phillips6  @alpertayfurr 100% agree sovereignty over your own context is non-negotiable for us at Seedling. That's why everything Bob learns about you stays on your machine by default. Your code, your behavioral profile, your conversation history all local in ~/.bob/.

Nothing leaves unless you explicitly choose to sync. Full transparency, full control, always.

Give it a spin: pnpm add -g @bobsworkshop/cli

#DreamBuildGrowTogether🌱

@kevinzrzgg Really appreciate this question these are exactly the failure cases we obsess over. Today, every file edit goes through an approval gate (nothing writes without your explicit yes), and every approved write auto-backs up the original to .bob-backups/ with a timestamp. Our auto-fix agent also runs programmatic validation before writing if the output looks corrupted, it's rejected and the file is never touched.

Honest gap: we don't have a dedicated bob rollback command yet, thanks to your questions that's now listed as something mandatory to be implemented starting today. Until then, recovery is .bob-backups/ or git checkout. Expect a proper rollback flow in the coming days. (we release updates EVERY day... 😊)

@fatih912 That's exactly why we built it this way. Your code, your conversations, your behavioral profile all stored locally in ~/.bob/ on your filesystem. Nothing leaves your machine unless you explicitly choose to sync. Full sovereignty by default, not by workaround.

Please let me know if you need anything I'm here to answer all questions

@olly_betts1 Exactly right, voice chatting really stood out when we were building as the thing that makes the exprience click. Glad you like it!

@igorgurovich Exactly right! Voice just 'feels' different even if the pure information being exchanged is the same. We invested a lot of time into getting turn-taking right, though it's a bit of an art as well, of course. If you pause, Warren will wait (up to a point). Overall, latency optimisation has been a big deal, we land at sub-second is what you need to make it usable (ie no point shipping if >1s delay). Once we got there, now it's just faster = better and we try to improve all the time (the ecosystem of voice AI infra is advancing so fast, we re-engineer our stack basically monthly).

@chiefkeith Thanks Alex! Spot on, plugging the enormous gap in DIY personal finance tools between simple budgeting mobile apps and regulated financial advice is exactly why we built Warren

@alex_mann7 Thank you for the kind words! Though, of course, Warren does not stray into regulated financial advice, it's much more a planning and guidance tool to help you feel more confident and informed to make your own decisions.

@thamibenjelloun That's the beauty - you just talk to it :)

@natalia_iankovych Absolutely, where would you want us to support?

@preston_przeszlo UK-only ! :) Where in Europe are you?

@mcarmonas Household features to bring all of your family's finances on the platform 😍

@dhiraj_patel5 It's only available and targeting the UK atm, but yes! that's exactly the idea and it handles it well for UK contexts. US soon 👀

@farrukh_butt1 Yes, this was a massive request from our users - clearly explain the "why" of the plan.

https://meetwarren.co.uk/publications/how-we-built-warrens-cashflow-engine - all the maths is handled via our deterministic engine that the user can use directly in the UI and that Warren can also plug into our of the box

@naresh_chandanbatve Thanks for pulling it! 😊

The target servers themselves don't need internet access. As long as your device can reach them over a VPN, private network, or local network, ShellMate can connect.

For a completely air-gapped setup with no access to the ShellMate API at all, that's not supported yet. Self-hosted support is something I'm actively exploring since it's a common requirement for security-focused teams.

@retain_dev The encrypted vault is synced through the server, so your devices don't need to be online at the same time. When you sign in on a new device, it downloads the latest encrypted vault automatically.

Updates are also synced across your connected devices and shared workspace members in real time, so everyone stays up to date without manual syncing.

The important part is that ShellMate only stores encrypted data. The actual decryption happens locally on your device, so the server never sees your plaintext credentials.
Security is one area where I'm probably overly cautious. I'm still actively refining this part of the architecture to make it as robust and bulletproof as possible while keeping the user experience simple.

@anand_thakkar1 That's a great question, and honestly one of the reasons I started building ShellMate.

Right now, ShellMate focuses on securely managing access, credentials, and collaboration around infrastructure. For key rotation, it doesn't automatically push changes to servers yet. My goal was to first solve the day-to-day pain of credential sprawl, sharing access safely, and keeping teams organized.

That said, automated key rotation and distribution is definitely on the roadmap. I'd love to get to a point where rotating access across environments becomes a few clicks instead of a manual process spread across dozens of servers.

I completely agree with you on the prod vs staging separation. Everyone knows it's the right thing to do, but once a team grows, it often turns into a mix of shared credentials, old keys, and tribal knowledge. Making those boundaries easier to enforce without adding operational overhead is something I'm thinking about a lot.

Out of curiosity, how are you handling key rotation today? Is it mostly manual, or are you using something like SSH certificates, Ansible, or another access management solution?

@anand_thakkar1 I just realized I missed part of your question! Within ShellMate, updates are synced automatically. If a credential, host, or permission changes, the update is pushed to all connected devices and shared workspace members in real time, so there's no need for anyone to manually sync.

@ianhxu Thanks Ian!

Yep, absolutely 😊

You can use ShellMate as a solo user without creating a team workspace. Your credentials and hosts can sync securely across your own devices, so if you're switching between multiple Macs (or other devices), everything stays in sync and ready to use.

Also, ShellMate is free forever for individual users. The only things that may be part of a paid plan in the future are optional AI features. Everything else is built to be useful from day one without needing a subscription.

One thing I cared about a lot was keeping the onboarding simple. I didn't want another tool that takes an hour to configure before it's useful. You can install it, add or import your hosts, and start connecting in just a few minutes.

Team workspaces are there when you need them, but they're completely optional.

Out of curiosity, what are you using today for SSH and credential management? I'd love to hear if there are any pain points or missing features that annoy you in your current workflow.

@ada_johnsen initial setup is super quick:

1. Connect a data source (database, warehouse, or one of over 750+ SaaS connectors)

2. Connect to your Slack workspace

3. Start asking questions

@farrukh_butt1 exactly! Some of our users use Basedash exclusively through Slack which is pretty cool.

@andika_fadhilah our AI builds up its own context layer based on your data schema, plus you can add additional context, skills, and deterministic metric definitions to improve consistency of answers.

@gabriella_anjani appreciate it, we agree!

@novamaker01 depends where it’s posted! If you want to keep it private you can use a private Slack channel or DM. You can choose to share that with other users if you like, regardless of their access level.

@scott_davidson_jr most common is connecting an existing database or warehouse, but it’s common to supplement that with data from Stripe, HubSpot, GitHub, Linear and tools like that.

@mcarmonas I built this for myself first! I record many musical ideas, but the applications span to creatives, podcasts, narrations, etc.

@farrukh_butt1 I have been toying with the idea of forking off ideas from a certain point. I guess I'm trying to figure out how to manage it. As a songwriter / producer, it's a problem that happens a lot 😅

I think a "source" version, and then deriviatives... hmmm haha. Well. there's my brain dump about the idea. It will be a thing though! for now, embrace the totally destructive nature of the app!

@alina_tyslenok_ Thank you! I have always wanted a simple layering tool, and then scope crept myself into what we see as Tide 😄

@varun1jan Thank you so much!

@daria_goroshko1 Hi Daria, great question! All document scanning and text recognition (OCR) happens 100% on-device using Apple's built-in Vision framework — nothing is ever sent to a server or third-party AI. The scanned images and extracted text are stored locally with full file-level encryption (iOS Complete File Protection). There's no cloud processing, no analytics, and no network calls at all.

@pranavshirole Awesome project, Pranav! The frustration of scattered medical records is incredibly relatable. I absolutely love the privacy-first approach and the one-time purchase model. We definitely need more apps built this way. Best of luck with the launch!

@pranavshirole Congrats on the launch, Pranav.

The idea immediately resonated with me because everyone has experienced that scramble of trying to find an old prescription, test result, or medication history when they actually need it. It's one of those problems that's surprisingly universal.

What I find particularly interesting is your decision to go all-in on privacy. In a world where most health products are built around collecting more data, building something offline, account-free, and owned by the user feels like a refreshing approach.

Wishing you the best with the launch. I'd love to learn more about the business side of Medicyn and how you're thinking about growth from here. What's the best email to reach you on? Happy to connect and exchange ideas.

@farrukh_butt1 Hi Farrukh, thank you for your comments. Yes! The Emergency Card has a "Share as PDF" option that generates a neatly formatted one-page summary (allergies, current medications, implants, emergency contacts, insurance info) and opens the iOS share sheet — so you can AirDrop, message, email or print. Takes seconds, no internet needed. In fact, each and every record can be exported in this way.

@igorgurovich Hi Igor, thank you for your comments. Good luck with your app, and I cannot wait to see what you build! Each record has a "Share as PDF" option that generates a neatly formatted PDF and opens the iOS share sheet — so you can AirDrop, message, email or print. Scanned documents can also be similarly shared in their original format.

Two things help here: (1) Medicyn supports multiple family profiles on the same device (e.g., a caregiver managing both their own and a parent's records in one app), and (2) you can export a full encrypted backup and share/import it on another family member's device if needed. It's manual rather than real-time, but it keeps your family's health data from ever touching a third-party server.

Right now, everything is local-first and lives on a single device — there's no cloud sync or live shared access between devices (by design, for privacy). Read-only shared access across devices is something I'm considering for the future, but it's a tradeoff against the zero-server privacy model I've built the app around.

@mcarmonas Thanks, Martí! 🙌

For now I'm focusing on a zero-friction onboarding experience.

Users install the extension and can immediately access launch velocity, upcoming launches, historical patterns and race dynamics without registration or setup.

At this stage I'd rather learn from real usage before introducing more onboarding steps or account-based features.

@busmark_w_nika Thanks, Nika! 🙌

I had the same feeling while researching the space. Most tools focus on launch day itself, while I wanted to provide more context around upcoming launches, launch dynamics and historical trends.

Really appreciate the feedback!

@thamibenjelloun Thanks! Great suggestion.

Historical benchmarking is something I'd love to add. Looking only at raw votes doesn't tell the whole story.

Comparing a launch against similar products by category, follower count, launch day competition and historical performance would give makers much more actionable insights.

Definitely on the roadmap.

@montverde Thanks, Luke! 🙌

That's exactly one of the motivations behind PH Radar.

The daily leaderboard is useful, but it naturally highlights the products with the most visibility and momentum. Many interesting launches never reach the top positions, yet they can still have strong engagement, unique ideas, or loyal communities.

That's why I wanted to include things like launch velocity, race dynamics, upcoming launches, historical patterns, and Community Spotlight — to help makers discover more of what's happening beyond the front page rankings.

Really appreciate the feedback!

@oshylabs Thanks, Arnold! 🙌

That's actually one of the directions I'm exploring. Right now PH Radar focuses on launch velocity, race dynamics, upcoming launches, historical patterns and watch rules, but comment activity is definitely an interesting signal.

I also agree that for many niche products, early discussion often tells a richer story than raw upvotes alone.

Good luck with your launch as well!

@farrukh_butt1 Thanks, Farrukh!

I had the same frustration when preparing launches. Choosing a launch date often feels like guesswork.

That's one of the reasons I added upcoming launches and momentum tracking.

Category-level attention before launch day is a great suggestion and definitely something I'll explore in future versions.

@tina_chhabra Exactly.

Launch velocity and race dynamics are the two things that usually decide whether a product finishes in the Top 5, Top 10, or disappears from the front page.

Most makers focus on launch day itself, but understanding the competition before launching is often just as important.

That's one of the main reasons I built PH Radar.

@gabriella_anjani Thank you 🙌

That's exactly what we kept running into. The words were technically correct, but the product just didn't feel like itself anymore.

Glad that resonated with you.

@farrukh_butt1 absolutely right. at some point every localization workflow seems to turn into "wait, where is the latest version of this file?" 😅 we're basically trying to make sure people never end up there in the first place

@ahmeterdempmk We definitely agree that blindly shipping AI-generated translations to production can be risky. That's why we have multiple review and validation layers that check translations, flag potential issues, and in some cases recommend manual review when confidence is low. We also support things like glossaries and terminology rules so teams can enforce consistency across languages. Our goal isn't to remove humans from the loop entirely, it's to dramatically reduce the amount of manual work while still keeping quality high.

@luca_capone That's exactly the kind of problem that got us interested in this space in the first place.

When the context is thin, we lean heavily on the tone profile, glossary, and the rest of the project content to build a better understanding of how the product communicates. A two-word button on its own is tough for any system, but it becomes a lot easier when it's translated as part of a larger product instead of as an isolated string.

We're definitely not claiming to solve context perfectly every time, but we've found that giving the model a stronger understanding of the product itself gets much better results than treating every string independently.

@novamaker01 It goes quite a bit further than basic validation. It'll catch things like missing keys, broken file structure and yes placeholder mismatches which will fail the check like you suggested. It also checks for untranslated content, makes sure HTML tags like <strong> or <br> are preserved, and warns about suspicious cases like where a translation is dramatically longer or shorter than the source. We also support things like the --min-coverage flag if you want to enforce a minimum translation coverage in CI. If you're curious, the docs are linked on our landing page. And if you end up trying it out, feel free to reach out with feedback or feature requests. We'd genuinely love to hear what would make it fit better into your workflow

@marianna_tymchuk Thank you! That's exactly what we kept running into. Getting a translation is easy now, but getting it to sound like the same product is a completely different challenge. Really appreciate the support 🚀

@itamarkenan 🏆🏆🏆

@farrukh_butt1 it is fully end to end customizeable, it can be per team, league, match or even a player. You chose what to be woken up for, just score margin, time, and a lot of advanced features like big chances burst, leading underdog and and more…

We have built a smart statistic model which gets the match stats and calculate by 1-10 how watch-worthy the match is (the “Drama Radar”).

Feel free to send me a message and talk about it 🙂

@mcarmonas I think the drama radar, in football it’s hard to tell wether the match is “clutch” or not.

For example 1-0 with the leading team dominates is not as clutch as 1-0 with the trailing team dominates. Building a model that knows how to handle those things.

@oshylabs thank you, penalty shootouts are a special case. You can choose to wake up only for them to

@louislecat You can wake up for goals, and some smart options can try to “predict” a goal is soon to come and wake you up.

Like big chances burst, the “Drama Radar” we built and a lot more.

Feel free to message me and we’ll talk about it🙂

@farrukh_butt1 Keep started as an iPhone project however I figured why not do both Mac and IOS! I agree though, I love getting some work done with my phone on a magsafe stand and a nice theme showing!

@marianna_tymchuk Appreciate it! they are my favourite themes too. If you could add a scene what would you like to see?

@montverde I like your thinking! Yes it could!

@margarita_s88 Huge congrats on the launch! 👏 Combining screenshot context with AI and human review is exactly what modern localization workflows need right now. what AI models are powering the analysis under the hood?

@margarita_s88 Dmitry Winicki Congrats on launching QACAT! Love seeing a tool built by people who live the problem daily. When you have a clear ICP, and clear pain, you get a clear fix. Curious - did this start as an internal tool at Alconost before you decided to make it a product?

@dnz_zh Thank you!

@derickd thanks a lot for your support Derick! I hope you'll find QACAT useful, and Dmitry will be happy to show a quick demo if you want to see it in action on your content! :)

@derickd Thank you, I really appreciate that!

I’d be happy to give your team a demo. Quality is ultimately about trust.

In our experience, not every piece of content needs the same level of review. For lower-risk content, fast and cost-effective checks like AQA, AIQA + human validation are often enough. But for sensitive content with a direct business, legal, or customer impact, deeper reviews such as LQA or LQT provide an additional layer of confidence.

Happy to show how these approaches can fit into your existing localization workflow.

@dioiv Thanks a lot. I love and miss Nitro.

@nickzaleski Thank for your inspiration, master :)

@kristina__grits We've kept plans affordable for both individuals and teams, you can check out pricing options here: https://qacat.alconost.com/pricing

@kristina__grits Thanks for asking. Good question. We’re still in the early stage, so pricing is intentionally simple and will likely evolve based on customer feedback and usage patterns.

@marcelo_macedo2 Yes! There is a free one-month trial, so you can run real projects through it before deciding 🙌

@olga_sapach Quick LLM QA checks are absolutely a valid modern approach, and I use them myself. The difference is that QACAT combines multiple quality layers rather than relying on a single AI opinion.

Depending on the workflow, that can include deterministic QA checks, + LLM-based evaluation, +human validation, and + deeper review workflows such as LQA or LQT.

In our experience, even the best and most expensive LLM models are excellent at finding some types of issues, but they still miss context, make incorrect assumptions, or generate false positives. That’s why we treat AI as one layer of the quality process rather than the final authority.

@margarita_tsygankova Thanks for asking. Good question indeed! We use two OCR engines. One supports Arabic, Hebrew (AI assisted), and other non-Latin scripts well (no AI involved, for NDA safe mode), while the other is optimized for Latin-based languages. We select the most suitable OCR approach based on the content being processed.

@kseniya_avtukhovich thanks for checking out QACAT! yepp, it is really convenient, our first users (linguists) already let us know they loved this specific feature ;)

@elena_kozak Thank you for your question - and yes! QACAT has already been used across 40+ languages in production environments. Not necessarily within a single project, but across multiple real-world localization workflows and quality evaluation scenarios.

@rohovets_taras Interesting concept. Turning funding announcements into detailed profiles with contacts, hiring signals, and market context could save a lot of research time for anyone doing outreach or market analysis.