[HN Gopher] Show HN: Dumped Wix for an AI Edge agent so I never ...
___________________________________________________________________
Show HN: Dumped Wix for an AI Edge agent so I never have to hire
junior staff
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
Author : axotopia
Score : 14 points
Date : 2026-03-19 15:59 UTC (7 hours ago)
| daveguy wrote:
| 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 wrote:
| I can't tell cringe AI-bubble business pivots from AI psychosis
| anymore
| axotopia wrote:
| 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 wrote:
| 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 wrote:
| 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 wrote:
| 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 wrote:
| 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.th
| eregister.com/2026/02/16/semantic_ablation_ai_...
| sonofhans wrote:
| > 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 wrote:
| 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 wrote:
| 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 wrote:
| 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 wrote:
| 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.
| lovich wrote:
| > 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.
|
| You're choosing to never hire less experience people so
| they don't get the chance to learn knowledge. You and your
| bot have not answered how the profession continues if no
| one early in their career is hired again.
|
| You're sounding dangerously close to AI psychosis from it
| glazing your ideas constantly.
| Rijanhastwoears wrote:
| > 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 wrote:
| I hope OP has friends.
| WarcrimeActual wrote:
| Shh. You're stepping all over his advertisement.
| sonofhans wrote:
| He's replying here in good faith and continuing the
| conversation. You cannot ask for more than that.
| linkjuice4all wrote:
| 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 wrote:
| 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.
| linkjuice4all wrote:
| I tried to give you the benefit of the doubt and tried to
| find some more info on building out a sub $1.5mm property in
| SoCal (feel free to pull/repost the logs). I'll admit that I
| did get some value out of the responses but it basically
| boiled down to {response from LLM based on "aggressive" sales
| prompt}{casual mention of mildly-related "deployment"}{push
| to contact form}.
|
| I work on the demand-side of marketing so I understand the
| frustrations of lead-gen but this didn't really seem to add
| any value beyond what ChatGPT et al would give you. Why
| should I give you my contact info?
| axotopia wrote:
| To answer your question, the contact info is for when the
| LLM reaches the end of its logic. I'm a 2-person firm,
| cannot compete with ChatGPT on general knowledge, but I'm
| also intentionally 'crippling' the bot's autonomy for
| liability reasons.
|
| It is designed to move the user toward a human once/if the
| technical complexity gets high, so it doesn't accidentally
| misrepresent a code or site issue, that could get us into
| legal trouble.
|
| It is a balancing act between utility and risk, still
| tuning where that line sits based on the stress test today
| , if that makes sense?
| albatross79 wrote:
| There is slop and there is cringe, you've managed to create both.
| discreteevent wrote:
| 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 wrote:
| 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 wrote:
| 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 wrote:
| 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.
| axotopia wrote:
| 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!
| pavel_lishin wrote:
| Some of your logs seem to include personal information about
| people.
| axotopia wrote:
| 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.
| unsnap_biceps wrote:
| at the bottom of your list, you have a number of bio links
| that I presume are you and your employees, but it's not clear
| that they are intended to be public from first glance.
| axotopia wrote:
| Good catch. Those bios and lists are essentially grounding
| truth for the training data for the RAG. Intent is twofold:
| Give bot truth anchors to stay focused and manage
| hallucination. Brain-dump for SEO crawlers to index our
| expertise. Removing the fluff from our site didn't agree
| with the 1999 SEO tech. I'm going to tighten the access and
| formatting so it doesn't look like an accidental exposure
| to a casual observer. Appreciate the sharp eye!
| csto12 wrote:
| This is clearly rage-bait. Why feed the troll by commenting?
| spzb wrote:
| Hope you like burning through tokens 'cause your bot has no
| restrictions at all. It just gave me Python code for calculating
| pi and then a detailed itinerary for a three day trip to
| Vancouver.
| axotopia wrote:
| haha, good catch! I focused on the AEC logic but did not put
| enough on the general guardrails. Your vancouver and Pi
| questions, including someone else on marine vessels seemed to
| have slipped thru. Got questions about refining Uranium and
| stuff, I'll be tightening the system prompt to manage this.
| Appreciate the stress test!
| stuaxo wrote:
| The linked chat log is unreadable because of the verbose AI slop.
|
| I say this as someone that has been building systems on LLMs and
| uses LLMs in dev work.
| doc_ick wrote:
| Are you worried about someone training their own model and
| recreating their own bot off of your public logs?
___________________________________________________________________
(page generated 2026-03-19 23:01 UTC)