I'm Yashar, founder of PivotPath, an AI product studio in Toronto. I spend my weeks helping organizations figure out where AI actually fits in how their teams work — and building the tools that follow. Most of what's written about AI is either hype or fear, and neither helps a leader decide what to fund next quarter. AI Signal is my attempt to cut through the noise: what happened, why it matters, and what I'd actually do about it.
―――――――――――
xAI’s Grok Voice Think Fast 2.0 improves speech-to-speech intelligence, transcription, conversational quality, tool use, and response speed. Priced at $0.08 per audio minute, it aims to make voice agents more reliable in real customer workflows.
The Signal: I would not put a human-like AI voice in front of customers yet: the risk is not just a wrong answer, but a wrong answer delivered with convincing confidence. Internal use cases, however, are ready for experimentation. Choose one low-risk workflow, build a prototype, and learn where the technology fails before deploying it in a customer-facing service.
―――――――――――
Google’s Lyria 3.5, now available in Flow Music, improves AI-generated songs through richer melodies, stronger lyrics, more expressive vocals, better pronunciation, and greater control over tempo and duration, giving creators more flexibility in shaping complete tracks.
The Signal: For individual creators, AI music is genuinely democratizing. For brands, however, it creates a familiar risk: the output is now good enough to use, but not distinctive enough to strengthen your identity. We saw the same cycle with generic AI-generated images: early use attracted attention, but widespread use quickly signalled low effort. Organizations should still experiment with these tools, particularly in low-risk settings such as internal training, prototype demonstrations, and pitch mock-ups. But customer-facing content should meet a higher standard.
―――――――――――
This article argues that increasingly capable AI systems may monetize each unit of compute so effectively that demand outpaces constrained supply, potentially driving compute prices sharply higher, strengthening frontier labs, and making lower-value AI applications uneconomical.
The Signal: This was the best thing I read this week. The exact increase in compute costs is uncertain, but the direction matters—and too few organizations are modelling it. If your AI business case assumes today’s per-token pricing will hold, test it again at 3x, 5x, and 10x. Which use cases still make financial sense? How dependent are you on a single vendor, and what would switching cost in engineering time? Will higher compute costs be passed to customers or absorbed into already-thin margins?
―――――――――――
DoorDash has received FAA Part 135 air carrier certification to launch DoorDash Air, an in-house commercial drone delivery operation. The company is developing its own aircraft and infrastructure, moving beyond partnerships toward greater control of autonomous last-mile logistics.
The Signal: The bigger story is not drones—it is vertical integration. DoorDash has decided that the delivery layer is too strategically important to keep renting from partners. Many companies are facing the same decision with AI: build critical capabilities internally or depend on platforms that may eventually compete with them. Drone delivery will likely scale more slowly than the headlines suggest. Airspace management, bad-weather reliability, liability, noise, and privacy are not simply engineering challenges; they require regulation, infrastructure, and public trust. But the underlying make-or-rent decision is already on business leaders’ desks today.
―――――――――――
That's this week's signal.
One ask, and I mean it: hit reply and tell me what you're actually seeing — the pilot that stalled, the tool nobody uses, the thing that worked. I read every one.
Talk soon,
Yashar