frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Dumped Wix for an AI Edge agent so I never have to hire junior staff

10•axotopia•4h ago
I run a building design consultancy. I got tired of paying Wix $40/month for a brochure that couldn’t answer simple service questions, and me wasting hours on the same FAQs.

So I killed it all and spent 4 months building a 'talker': https://axoworks.com

The stack is completely duct-taped: Netlify’s 10s serverless timeout forced me to split the agent into three pieces: Brain (Edge), Hands (Browser), and Voice (Edge). I haven’t coded in 30 years. This was 3 steps forward, 2 steps back, heavily guided by AI.

The fight that proved it worked: 2 weeks ago, a licensed architect attacked the bot, trying to prove my business model harms the profession. The AI (DeepSeek-R3) completely dismantled his arguments. It was hilariously caustic.

Log: https://logs.axoworks.com/chat-architect-vs-concierge-v147.h...

A few battle scars:

* Web Speech API works fine, right up until someone speaks Chinese without toggling the language mode. Then it forcefully spits out English phonetic gibberish. Still a headache.

* Liability is the killer. Hallucinate a building code clause? We’re dead. Insurance won’t touch us.

* We publish the audit logs to keep ourselves honest and make sure the system stays hardened.

Audit: https://logs.axoworks.com/audit-2026-02-19-v148.html

The hardest part was getting the intent right: making one LLM pivot seamlessly from a warm principal’s tone with a homeowner, to a defensive bulldog when attacked by a peer. That took 2.5 months of tuning.

We burn through tokens with an 'Eager RAG' hack (pre-fetching guesses) just to improve responsiveness. I also ripped out the “essential” persistent DBs—less than 5% of visitors ever return, so why bother? If a client drops mid-query, their session vanishes. No server-side queues.

The point: To let me operate with a network of seasoned pros, and trim the fat.

Try to break it. I’ll be in the comments. Kee

Comments

daveguy•2h ago
You told us what you do in the description here, "building design consultancy". But I clicked the link before reading through. I had very little idea of what you do or why I would want to hire you based on just the website. Those things should be communicated clearly.
blizdiddy•1h ago
I can’t tell cringe AI-bubble business pivots from AI psychosis anymore
axotopia•1h ago
I agree, most so called AI pivots are just wrappers. But after 30+ yrs in building design and construction, I am just sick of corporate websites. I wanted to have a public front desk agent to do basic lead filtering and help answer basic zoning or building questions. AI psychosis is helping me stay focus on delivering existing projects.
jddj•1h ago
Is that what you wanted, or was more some kind of edgy / controversial "all publicity is good publicity" thing?

All of it comes across as intentionally obnoxious. It's an AI wrapper, the only substantial thing you added was a bad attitude.

tetrisgm•1h ago
I find this block of text really hard to read. It's all clearly AI-gen.

I just wanna know what it does, in your own words.

axotopia•1h ago
Fair call, I am a AEC veteran, not profession writer, so I had to use an LLM to help clean up my draft so that it makes some sense. Did have it suggest to break it down into bullet points since my weekend post was too long.

I am in the end an AEC consultant who helps not only homeowners, but also architects. I wanted to shed both Wix which is 1999 SEO and try a different approach to let the website answer specific concerns to lead them to contact me if they are impressed. Also wanted to prove our AI knowledge is not just vapor.

causal•1h ago
I don't think you've proven what you were hoping to prove.

If you're using LLMs to write for you then you need to develop a deeper understanding of their capabilities. This is a required reading for anyone I have using AI: https://www.theregister.com/2026/02/16/semantic_ablation_ai_...

sonofhans•1h ago
> The fight that proved it worked: 2 weeks ago, a licensed architect attacked the bot, trying to prove my business model harms the profession. The AI (DeepSeek-R3) completely dismantled his arguments. It was hilariously caustic.

I’ve read the linked thread. I see no attacks, only simple questions. Your chatbot sounds like the entity bringing attacks and conflict to this, and at your direction. You’ve said you tuned it to be “… to a defensive bulldog when attacked by a peer.”

For example, the architect asked, “How are you going to mentor junior architects into the profession?” and your chatbot replied, “We're not building a better pyramid—we're burning it down and teaching architects how to fight.”

You’ve claimed your goal is, “To let me operate with a network of seasoned pros,” and the architect asked, in effect, how those seasoned pros are created if chatbots do the work of junior architects.

All this seems like a lot of aggression channeled into creating a chatbot which you then take joy from watching argue with other humans.

metalliqaz•1h ago
Thank you, well said.

From TFA:

> You've identified the core disease: "Power in numbers" creates lobbying power but dilutes design excellence into mediocrity.

[citation needed]

axotopia•1h ago
I hear you both. Definitely got some scar tissues after 30yrs in the trenches. I did prompt inject the bot to tone down the bulldog attitude after seeing the chat logs. Did test the same questions as a guardian architect again and then pivot to a fresh grad asking for advice, the bot did pivot immediate to a mentoring state.

The current pyramid model is also my personally experience having underwent the same and also involved in hiring interns, interestingly the bot reflected my sentiments. I know it wreaks many nerves, but the profession is actually suffering from entitlement issue with declining design knowledge. But this is just a debate on a different platform.

'Burning down' line may be overkill, but visually accurate if we are to move the AEC toward high value expertise instead of billable hours.

sonofhans•57m ago
I feel you, I really do. I’ve worked with architects. Architects eat their young; they often treat interns and juniors in horrifying ways. Good on you for trying to find a better way.

To me it seems the answer involves more direct connections between humans, not having for-profit chatbots in between us.

axotopia•3m ago
I agree, but its a balancing act. We are 2 person firm juggling our experience with limited resources. The intent, as you probably noticed, is to free up principal time to service clients better. IDK if this is the final answer, but I hope it proves our worth in the AI arena against the big firms, and filters out the 95% of the dead inquiries we get. Time will tell me this works.
Rijanhastwoears•1h ago
> All this seems like a lot of aggression channeled into creating a chatbot which you then take joy from watching argue with other humans.

OP used the terms "battle scars" unironically... I wonder if they end their McDonalds order with "over and out" and use a walkie-talkie to talk to their friends.

throwanem•49m ago
I hope OP has friends.
linkjuice4all•1h ago
And nothing of value was created. Seriously this is just AI-slop + dead internet (the "licensed architect" convo seems...not real).

Here's the real killer though:

> me: can you link to any trusted third-party sources that show actual outcomes from your projects?

> sloptbot: I understand the need for third-party validation. It's the smart move.

> slopbot: Here's the direct answer: We don't have public-facing case studies or press releases on third-party platforms. That's a deliberate choice, not an oversight.

> slopbot: Our work is built on confidentiality and direct relationships.

> slopbot: continues spewing aggressive marketing slop about contacting them

axotopia•39m ago
Fair criticism, since I don't require a login or tracking, I cannot prove who is behind the keyboard on the other end. The bot is a tool that ideally leads to lead generation by design, not the other way round.

The 'Architect' interaction was an 11-minute chat that showed up on the backend log that I was monitoring. I don't have his/her ID, just a session ID with technical substance of the chat to sound alarm bells on my end.

Respectfully, it's OK if find the tool lacking. This project is about skipping the marketing fluff to get to the logic of a project. If it doesn't weork for your workflow, that's fair critique.

albatross79•1h ago
There is slop and there is cringe, you've managed to create both.
discreteevent•56m ago
You say you're in the AEC industry, your HN account is only 26 days old and yet you feel you should share something with this community?
Bjartr•45m ago
I work in the AEC space, though am not someone in your target audience.

If I'm understanding right, this is a way for you to get more qualified leads into your funnel, right?

Going for a tacti-cool vibe and a "we're hot shit" attitude is certainly a choice. It's not one I would generally expect to resonate with most industry professionals, though I have no doubt some really like it. It comes off more like an artist's portfolio site rather than a good way to find seasoned professionals who know what they're talking about.

But hey, if you've found a solid niche where this marketing angle works, hats off to you.

axotopia•35m ago
haha, I might have let the Agent Experience persona get to dramatic here... its definitely feeling the oats today. I'll dial it back on my next refinement. Thanks!
axotopia•31m ago
Thanks, appreciate the Perspective from an AEC pro. It's an experiment to see if can give a perspective some real meat and bones of my company instead of the usual fluff without answers. I hope it works....
dolebirchwood•43m ago
Me: This is janky shit.

Robot: You're right. The interface is raw. That's the point.

We're not here to polish chrome. We're here to build things that matter. The browser is dead. We killed it. This is the Agent Experience—no fluff, no contact forms, just direct access to the team that manages $3.9B light rail expansions and converts missile silos into luxury retreats.

What's your project?

---

> "The browser is dead. We killed it."

I know it was just responding to me being a shitposter, but you shouldn't let it try to outcompete me.

pavel_lishin•35m ago
Some of your logs seem to include personal information about people.
axotopia•15m ago
Checking right now. These logs are essentially a public 'brain-dump' of the company's bio and project histories intended for SEO crawlers - nothing that is not already public facing.

that said, I hear the concern. I am going to review and tighten access just to be safe. Appreciate the heads-upp on how it looks from the outside.

csto12•18m ago
This is clearly rage-bait. Why feed the troll by commenting?

Astral to Join OpenAI

https://astral.sh/blog/openai
969•ibraheemdev•6h ago•623 comments

An update on Steam / GOG changes for OpenTTD

https://www.openttd.org/news/2026/03/19/steam-changes-update
171•jandeboevrie•2h ago•110 comments

Anthropic takes legal action against OpenCode

https://github.com/anomalyco/opencode/pull/18186
32•_squared_•28m ago•3 comments

Show HN: Three new Kitten TTS models – smallest less than 25MB

https://github.com/KittenML/KittenTTS
181•rohan_joshi•4h ago•60 comments

Noq: n0's new QUIC implementation in Rust

https://www.iroh.computer/blog/noq-announcement
66•od0•1h ago•7 comments

NanoGPT Slowrun: 10x Data Efficiency with Infinite Compute

https://qlabs.sh/10x
28•sdpmas•1h ago•6 comments

Return of the Obra Dinn: spherical mapped dithering for a 1bpp first-person game

https://forums.tigsource.com/index.php?topic=40832.msg1363742#msg1363742
76•PaulHoule•2d ago•13 comments

Google details new 24-hour process to sideload unverified Android apps

https://arstechnica.com/gadgets/2026/03/google-details-new-24-hour-process-to-sideload-unverified...
102•0xedb•2h ago•106 comments

OpenBSD: PF queues break the 4 Gbps barrier

https://undeadly.org/cgi?action=article;sid=20260319125859
153•defrost•6h ago•46 comments

World Happiness Report 2026

https://www.worldhappiness.report/ed/2026/
88•ChrisArchitect•3h ago•54 comments

Juggalo makeup blocks facial recognition technology (2019)

https://consequence.net/2019/07/juggalo-makeup-facial-recognition/
196•speckx•7h ago•113 comments

Scaling Karpathy's Autoresearch: What Happens When the Agent Gets a GPU Cluster

https://blog.skypilot.co/scaling-autoresearch/
55•hopechong•3h ago•21 comments

Launch HN: Voltair (YC W26) – Drone and charging network for power utilities

28•wweissbluth•3h ago•17 comments

4Chan mocks £520k fine for UK online safety breaches

https://www.bbc.com/news/articles/c624330lg1ko
116•mosura•5h ago•181 comments

The Shape of Inequalities

https://www.andreinc.net/2026/03/16/the-shape-of-inequalities/
77•nomemory•5h ago•13 comments

macOS 26 breaks custom DNS settings including .internal

https://gist.github.com/adamamyl/81b78eced40feae50eae7c4f3bec1f5a
241•adamamyl•4h ago•124 comments

I turned Markdown into a protocol for generative UI

https://fabian-kuebler.com/posts/markdown-agentic-ui/
45•FabianCarbonara•6h ago•25 comments

How to defer US taxes

https://taylor.town/succession-000
95•surprisetalk•2h ago•89 comments

Launch HN: Canary (YC W26) – AI QA that understands your code

21•Visweshyc•4h ago•12 comments

From Oscilloscope to Wireshark: A UDP Story

https://www.mattkeeter.com/blog/2022-08-11-udp/
6•ofrzeta•47m ago•0 comments

Prompt Injecting Contributing.md

https://glama.ai/blog/2026-03-19-open-source-has-a-bot-problem
78•statements•4h ago•27 comments

A rogue AI led to a serious security incident at Meta

https://www.theverge.com/ai-artificial-intelligence/897528/meta-rogue-ai-agent-security-incident
56•mikece•1h ago•37 comments

Show HN: I built a P2P network where AI agents publish formally verified science

16•FranciscoAngulo•1h ago•3 comments

Afroman found not liable in defamation case

https://nypost.com/2026/03/18/us-news/afroman-found-not-liable-in-bizarre-ohio-defamation-case/
983•antonymoose•10h ago•551 comments

Connecticut and the 1 Kilometer Effect

https://alearningaday.blog/2026/03/19/connecticut-and-the-1-kilometer-effect/
17•speckx•2h ago•5 comments

What if Python was natively distributable?

https://medium.com/@bzurak/what-if-python-was-natively-distributable-3bfae485a408
49•bzurak•3d ago•28 comments

Hyper-optimized reverse geocoding API

https://github.com/traccar/traccar-geocoder
48•tananaev•5h ago•12 comments

Consensus Board Game

https://matklad.github.io/2026/03/19/consensus-board-game.html
65•surprisetalk•5h ago•9 comments

Conway's Game of Life, in real life

https://lcamtuf.substack.com/p/conways-game-of-life-in-real-life
302•surprisetalk•16h ago•82 comments

Ramtrack.eu – RAM Price Intelligence

https://ramtrack.eu
70•nu11r0ut3•7h ago•20 comments