Feed aggregator

Show HN: Conkoa AI – Voice-First Slack for Construction

Hacker News - Tue, 03/10/2026 - 12:54pm

I spent years building a SaaS product for the construction trades focused on health and safety. We obsessed over UI/UX because our users varied wildly in age, tech comfort, literacy, and preferred language. As good as we were, it was always a challenge to get new users to become proficient with it. That company was acquired in 2022.

One thing I noticed throughout that journey: internally, we lived in Slack. It became our hub for everything — team conversations, server alerts, support tickets. It was always the open tab. So naturally, we tried to get our construction clients to adopt it too.

None of them did. The reason was obvious in hindsight: Slack is built around typing. That doesn't work when you're on a roof, driving between jobs, or on a scaffold installing wall panels.

After some time off, I kept thinking about that gap. What if field teams had one unified interface where they could:

- Communicate with the rest of their team - Update their work orders, timekeeping, or other systems they already use - Ask questions answered by an AI trained on their company's own documents and data

All by voice. Any language. No typing. No learning multiple apps.

That's Conkoa. A foreman on a ladder can say "We had three workers on Level 3 for eight hours at the Green Cedar project" and that information gets properly entered into Procore or whatever their system of record is. Messages are automatically transcribed and translated, so a Spanish-speaking crew member and an English-speaking PM stay in sync without anyone doing extra work. Teams can upload photos and documents that feed into a RAG pipeline, so the built-in AI agent can summarize and surface that information on demand.

We integrate with tools like Procore today and are building out more integrations.

Conkoa is live and being used by dozens of companies across the construction trades. You can set up a free workspace at https://conkoa.ai — would love feedback from the HN community.

Comments URL: https://news.ycombinator.com/item?id=47325824

Points: 2

# Comments: 0

Categories: Hacker News

Ask HN: How will AI-generated code be vetted?

Hacker News - Tue, 03/10/2026 - 12:51pm

Obviously something the entire industry is struggling with, but curious about HN's thoughts on what a post-AI-coding SDLC looks like 2 years from now.

Specifically around "review", change control, and SRE.

Comments URL: https://news.ycombinator.com/item?id=47325791

Points: 2

# Comments: 0

Categories: Hacker News

You Can Still Get Windows 10 Security Updates for Free—Here's the Trick

CNET Feed - Tue, 03/10/2026 - 12:44pm
Windows 10 support ended back in October. Here's how to keep access to Windows 10 security updates without spending a dime.
Categories: CNET

Ask HN: How are you testing AI agents before shipping to production?

Hacker News - Tue, 03/10/2026 - 12:04pm

We've been running reliability audits on AI agents before production deployment, and the failure patterns are consistent enough that I built a framework around them.

Some context on why this matters right now: Gartner predicted over 40% of AI agent projects will fail by 2027. In January 2026, a prompt injection in a customer support agent processed a $47,000 fraudulent refund. These aren't fringe cases anymore.

The 7 failure modes we see most often:

1. Hallucination under unexpected inputs — works perfectly in demos, invents data when the input is slightly off

2. Edge case collapse — null values, Unicode names (O'Brien, José, 北京), empty fields, concurrent requests

3. Prompt injection — if your agent processes external content, users can hijack its behavior through that content

4. Context limit surprises — agent works for 95% of conversations, then silently misbehaves when the context window fills. No error. Just wrong behavior.

5. Cascade failures — tool call #1 fails, agent keeps going, by the time a human sees the result 3 calls have compounded the error

6. Data integration drift — built against your schema in January, schema changed in February, still calling deprecated endpoints in March

7. Authorization confusion — multi-tenant system, cached context from User A bleeds into User B's session

We've built 50+ test cases across these categories. Most teams test #1 and #3. Almost no one systematically tests #4, #5, and #6 before shipping.

Happy to share the framework. Curious what failure modes you've hit that I haven't listed.

Comments URL: https://news.ycombinator.com/item?id=47325105

Points: 1

# Comments: 0

Categories: Hacker News

Show HN: CocoIndex-Code open source embedded AST-based code MCP

Hacker News - Tue, 03/10/2026 - 12:03pm

Hi HN friends, George here. I've been working on an embedded AST based code mcp that is super light weight and just works. I published my test on the repo, it has great effect on achieving and speeding up coding tasks and saves tokens.

Works with Claude, Codex, OpenCode and any coding agents, I'd love to learn from your feedback!

Comments URL: https://news.ycombinator.com/item?id=47325095

Points: 1

# Comments: 0

Categories: Hacker News

Parseword – new game from the creator of Wordle

Hacker News - Tue, 03/10/2026 - 12:00pm

Article URL: https://www.parseword.com

Comments URL: https://news.ycombinator.com/item?id=47325050

Points: 1

# Comments: 0

Categories: Hacker News

Pages