👻Our voice AI scary stories continue with the phantom transfer: the caller hangs up a split second before the transfer completes, so a live agent answers a call with no caller, no audio, no context. Just a confused "Hello?" into the void. https://t.co/NMaVTgUggG
The AI gold rush is Act One, not the whole play. Gold rushes consolidate, and the value moves up the stack. Our CEO's ClueCon keynote on why a generation of builders is rebuilding the bottom of the stack and running out of money before shipping the top: https://t.co/nZbCHfIoC2
An AI voice horror story: two valid events hit the same call at the same instant, both read the same state, and the last write silently erases the other. Nothing crashes. Your agent just forgets what the caller said. Learn about the double update: https://t.co/jdGCETtbBF
It's that time of year. Black Friday texts hit a 98% open rate, which is why SMS is the channel you don't want to skip this season. But carriers can take days to approve a campaign. Get registered now, not the week of. Full playbook here: https://t.co/J8M71HCA5s
Running voter outreach, fundraising, or GOTV texts this cycle? Every political SMS campaign has to be registered with The Campaign Registry first, and political is vetted more strictly than most. A full resource hub to get approved on the first try: https://t.co/ReAI6amWxJ
🧟 Spooky season calls for a ghost story: the caller hung up, the media stream dropped, the PSTN leg is gone, but your system still swears the call is active. Billing never stops. Cleanup never fires. A dead call that won't die... https://t.co/sK49aAdyGC
Your customer chats with your bot on the website, then calls in and starts over from zero. It feels like contacting two different companies. AI Chat is one agent that follows the customer across voice and text without losing context: https://t.co/QPd8T0SM53
It's election season, and political SMS is ramping up. The catch: the political use case needs a Campaign Verify token, an EIN, and a compliant website before carriers approve you, and that token takes time to get. Start now. Here's what you need: https://t.co/kSixpcfU5d
Every serious communications app hides the same secret: beneath the product, there's a layer of state management, routing logic, and escalation handling that someone had to build from scratch. For over a decade, that layer wasn't a product. Now it is. https://t.co/sYPNU73BDP
Press 1 for billing. Press 2 for support. Press 0 to give up entirely. IVRs route. IVAs resolve. Here's what that difference actually means in production.
https://t.co/2KFrmYjh1V
Today's the day, Palo Alto. This afternoon you'll build an AI voice agent that holds up under real call volume, not one that only survives the demo.
You'll put the model in charge of language while your code owns everything that matters. Sign up:
https://t.co/U2darJEGpG
Tomorrow, Palo Alto. We're running a hands-on workshop on building AI voice agents that don't fall apart in production.
You leave with an AI phone agent where the model handles the conversation and your code handles the logic. https://t.co/ZRTAojZArL
"Never offer a refund we don't have." Your prompt says it. The model does it anyway.
The fix isn't a firmer prompt. It's scoping.
At our Palo Alto workshop on Tuesday, October 6, we build agents where the model's reach is restricted by design.
Register: https://t.co/DsXMViWjM2
You can segment your audience based on RCS delivery receipts, sending RCS to supported devices and SMS to everyone else. That's just one of the practical questions answered in this Q&A with SignalWire's carrier ops team: https://t.co/15RkrM9ZQj
Palo Alto, we're coming to you.
If you've ever shipped a voice agent that demoed clean and then set off firedrills in prod, this one's for you. One evening, hands on keyboards, building agents that behave the same at 10 calls as they do at 10,000. Oct 6th. https://t.co/scdSgq0EOF
Today's the day. San Francisco, we're on.
Build AI voice agents that don't break in production. Tonight, 5:30 PM.
Bring a laptop. Bring a charger. That's it.
https://t.co/lzXxHwnldU
See you tonight.
#voiceai#sanfrancisco
The ClueCon 2026 Videos are Here! All 32 sessions from Chicago are now live on the FreeSWITCH YouTube channel. Three days of real-time comms, carrier-grade voice AI, security, WebRTC, and the unglamorous parts of getting agents into production. Watch now: https://t.co/wLGC8903OF
TOMORROW: We're coming to San Francisco for a free hands-on voice AI building workshop.
You walk out with:
- A working voice agent
- An AI-driven IVR
- Function fillers to kill dead air
- An open-source observability dashboard
Register: https://t.co/PM7aUzPXWl
A prompt is a suggestion, not a constraint. It will follow your rules most of the time. But “most of the time” is not an engineering standard. Especially not for a contact center handling thousands of calls a day.
https://t.co/67Fz4Rmqot
Voice AI doesn't fail because the model is bad. It fails because the model is driving. Here's the architecture shift that makes voice AI behave the same on call #10,000 as it did on call #1: https://t.co/z7jvIZFh4N