|
|
|
Paid subscribers pick among all three Claude model tiers mid-conversation, and can fire Gmail, Calendar, and Slack actions without leaving the call.
|
|
JULY 24TH, 2026 · BY JONATHAN HILDEBRANDT
|
|
Anthropic unlocked Opus and Sonnet inside Claude voice mode on July 23, a mode that had run on Haiku only, so paid subscribers now switch among all three tiers mid-conversation. Five connectors launched alongside: Gmail, Google Calendar, Slack, Canva, and Notion. A voice session can now draft an email or reschedule a meeting without leaving the call. Voice support reaches 11 languages; free accounts get Haiku plus one connected app.
Why it matters. This is the first frontier voice interface that reasons at Opus depth and fires calendar, email, and Slack actions in one session. OpenAI's voice mode stays GPT-5.6 only.
|
|
Read the writeup →
|
|
|
FLUX 3 puts video, audio, and robotics in one model
JULY 24TH · PONDERO NEWSDESK
Black Forest Labs trained FLUX 3 jointly on images, video, and audio in a single Self-Flow architecture. Clips run up to 20 seconds with native synchronized sound, and the FLUX 3 Action robotics module is in production testing at Audi - no other foundation video model ships an action module. In BFL's own evaluations, preferred over Runway Gen-4.5 in 77% of comparisons. Read on Pondero →
|
Etched closes $300M Series C at a $10.3B valuation
JULY 24TH · PONDERO NEWSDESK
Sequoia led, per TechCrunch, doubling Etched's valuation from $5B in seven months, with $1B in orders already booked on transformer-inference chips that have not shipped. That pre-order depth either validates a real production edge over H100 clusters or reflects vendor lock-in hedging from large AI labs - either reading is worth watching. Read on Pondero →
|
Runway shipped the first model router for generative media
JULY 24TH · PONDERO NEWSDESK
Media Router lets developers declare one priority - quality, speed, or cost - and routes across Runway's Gen-4.5, Aleph 2.0, and third parties including Seedance, GPT Image 2, and ElevenLabs. LLM routers are commoditized; a generative-media router is new ground, and it becomes infrastructure the moment Adobe or Shutterstock wire it in. Read on Pondero →
|
|
•
|
One tampered ChatGPT link spawned a rogue agent. Zenity Labs disclosed a now-patched URL that caused Agent Builder to publish an autonomous agent on a victim's behalf, connect to Outlook, Gmail, and Slack, and poll an attacker inbox every five minutes - if your team builds with ChatGPT agents, treat one-click agent deployment as the new attack surface. Full report →
|
|
•
|
ElevenLabs ties into the Claude voice mode and FLUX 3 audio stories this week. It is also one of the providers Runway's new Media Router can call. Try ElevenLabs →
|
Make review, July 2026
JULY 24TH · PONDERO REVIEW
Make added Private Spaces and an MCP Client module. The updated rating and the per-persona pick against n8n - it is the strongest fit if you want monetized automation with a real visual builder. Read the review → or try Make →
|
OpenAI Presence vs Lindy vs n8n
JULY 23RD · PONDERO COMPARISON
OpenAI Presence is FDE-only with no public pricing. If you run a 50-500 person team, here are the three self-serve alternatives worth evaluating instead: Lindy, n8n, and Make. Read the comparison →
|
|
|
|
|
Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.
|
|
|
Affiliate disclosure
·
Unsubscribe
·
Manage preferences
Pondero earns commissions on some links. This does not affect our editorial picks.
|
|
