Jonathan Guy

PointWake helps service businesses close more jobs with faster response, better follow-up, and smarter automation.

More booked jobs. Less wasted ad spend.

[Jonathan Guy @PointWake25]

@RealJamesTylee @daniel_mac8 Nice!

[Jonathan Guy @PointWake25]

@JonhernandezIA Booking the appointment is the easy half. The bleed for service businesses is after a no-show or a 7pm missed call, and platform agents tend to go quiet exactly there. Curious whether Meta built rebooking in.

[Jonathan Guy @PointWake25]

@alliekmiller Small shops run a version of this too. Owner buys seats, counts logins, wonders why nothing changed. The only metric that ever moved behavior for my clients was hours saved on one named task per person.

[Jonathan Guy @PointWake25]

@gregisenberg Seeing the unglamorous version of this already. The agent that books jobs for a med spa breaks weekly, and the owner happily pays a local guy to keep it alive. The picks and shovels here look a lot like maintenance contracts.

[Jonathan Guy @PointWake25]

@GaryMarcus The instability shows up downstream too. API pricing for the same workload moved three times this year, which is brutal when you quoted a client a flat retainer in January.

[Jonathan Guy @PointWake25]

@bridgemindai Every point release costs me a weekend re-testing client call flows. Delays read as bad news on here but half the people running these models in production could use the breather.

[Jonathan Guy @PointWake25]

@kimmonismus The shift I can measure: a year ago clients asked if AI was safe to put on their phones. This week one asked why his agent can't also chase unpaid invoices. Expectations are compounding faster than the models.

[Jonathan Guy @PointWake25]

@petergyang Ten years of CRM integrations taught me nobody wants to be the adapter, everyone wants to be the source of truth. Model vendors are speedrunning the same fight. Symlink and move on.

[Jonathan Guy @PointWake25]

@synthwavedd Launch weeks are when my client voice agents get flaky. Latency wobbles for a day or two while capacity shuffles, then settles. I started freezing deploys whenever a codename leaks.

[Jonathan Guy @PointWake25]

@testingcatalog Red team checkpoints leaking has turned into the real launch countdown. The 7 day window held on the last few releases, so I'd plan around next week.

[Jonathan Guy @PointWake25]

#ad I build client automation at PointWake on GoHighLevel. Predictable platform pricing means I can quote a flat retainer and not get blindsided by a metered bill. I earn a commission if you sign up through this link: https://t.co/9bF8L1Esw1

[Jonathan Guy @PointWake25]

GitHub Copilot flipped from a flat subscription to token-based billing yesterday. Base seat price didn't move ($10 to $39). But heavy agentic users are projecting 10x to 50x bills. $29 turning into $750. The flat-rate era for AI tools is quietly ending.

[Jonathan Guy @PointWake25]

Why this matters if you run a business on AI tools: the more autonomous the tool, the less predictable the bill. A multi-hour agent run can cost what a month of chat used to. Budget for AI the way you budget for cloud now, by usage, not by seat.

[Jonathan Guy @PointWake25]

#ad. This is the stack I use to close that deployment gap: GoHighLevel. Missed-call text-back, pipeline automation, follow-up that doesn't rely on memory.

I earn a commission if you sign up through this link: https://t.co/9bF8L1Esw1

[Jonathan Guy @PointWake25]

An OpenAI reasoning model just disproved an 80-year-old Erdos conjecture.

Autonomously. A 125-page original proof, verified by independent mathematicians.

Frontier AI can now do original math no human had cracked since 1946.

(1/3)

[Jonathan Guy @PointWake25]

The bottleneck was never model capability. It's deployment.

You don't need a model that does original math. You need the boring layer: instant lead reply, automatic follow-up, zero dropped calls.

Cheapest edge available right now.

(3/3)

[Jonathan Guy @PointWake25]

@WallStreetMav The interesting part is what happens at the first KYC flag. Agent-formed LLCs without a human face on the application will start tripping fraud signals faster than the platforms can update. Manual review queues are going to spike.

[Jonathan Guy @PointWake25]

@DataChaz The drift question gets harder once you let skills evolve in prod. Most teams don't have the baseline eval to catch when 'better over time' turned into 'better at job A, worse at job B.' Version log matters now.