Google ships webhooks in Gemini API for long-running tasks
Today's top 24 insights for PM Builders, ranked by relevance from X, Blogs, YouTube, and LinkedIn.
Google ships webhooks in Gemini API for long-running tasks
#1 𝕏
xAI launched emotion-rich voice cloning on its Grok Voice API, now live for developers to generate AI voices nearly indistinguishable from human speech.
#2 𝕏
Logan Kilpatrick shipped Webhooks in the Gemini API to streamline developer workflows for long-running tasks like batch jobs, agents, and GenMedia.
#3 𝕏
NVIDIA AI launched cuOpt Agent Skills, delivering GPU-accelerated decision optimization for supply-chain planning. First 50 developers who deploy the launchable on NVIDIA Launchable get free credits.
#4 𝕏
NVIDIA AI now offers end-to-end support in Megatron Core for training 30B-scale Kimi K2 and Qwen3 models with higher-order optimizers (Muon, MOP, REKLS), pushing efficiency on GB300 GPUs and NVL72 systems beyond standard data-parallel methods.
#5 📝 Simon Willison
Granite 4.1 3B SVG Pelican Gallery - Simon tried prompting different quantized variants of IBM's Granite 4.1 3B model to 'Generate an SVG of a pelican riding a bicycle' and published a gallery of the results. He found no clear relationship between model size and output quality — most results were poor.
#6 📝 Anthropic News
Building a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs - Anthropic announced plans to build a new enterprise AI services company in partnership with Blackstone, Hellman & Friedman, and Goldman Sachs.
#7 📝 OpenAI News
How OpenAI delivers low-latency voice AI at scale - OpenAI explains engineering techniques to achieve low-latency voice AI at scale, covering system design, model optimizations, and infrastructure approaches. The post outlines how these measures reduce end-to-end latency for real-time voice applications.
#8 𝕏
Logan Kilpatrick rolled out extensive error message improvements for the Interactions API, making its feedback far more human- and agent-readable.
#9 𝕏
Philipp Schmid launched a QoL upgrade for the Interactions API: errors now name the exact field and bad value, list supported enum options, show expected vs. actual formats, and pinpoint field paths like `input[0].name`.
#10 𝕏
Guillermo Rauch launched npx deepspec, an open-source agent orchestrator that leverages thousands of parallel coding agents in Vercel Sandbox to uncover critical security vulnerabilities in minutes.
#11 ▶️
AI Agents run my business and life
Greg Isenberg
Andrew Wilkinson demonstrates running an autonomous SaaS business using Claude-based OpenClaw agents orchestrated in Harbor to handle support ticket triage (including auto-fixing P0 issues and merging PRs) and marketing campaigns via Post Hog and Meta/Reddit ads.
- The marketing agent hooked into Post Hog performs multivariate testing, generates ad creative, sets budgets (scaling by $1,000 at a time) and has driven $20,000 of revenue.
- A Pinecone vector database on his family office Folly Partners answered “break down how many minority venture investments I’ve made,” reporting 132 direct investments with $16 million invested and $36 million current value.
- His CFO, without prior coding experience, built in two weeks a custom portfolio-tracking app using vibe coding and Cloud Code that replaces a $50,000–$100,000/year Adapar subscription by piping into banking, accounting, and market data.
#12 in
Claire Vo: Owen Williams at Stripe built Protodash, an AI prototyping studio on dev boxes with Cursor + CC that spins up real, clickable prototypes in minutes—no more “blurple slop.
#13 𝕏
Garry Tan announced GStack v1.26.3 integrates with GBrain, boosting coding agents with richer context than before.
#14 📝 Jesse Vincent
Superpowers 5.1.0 - Announcement of Superpowers 5.1.0, highlighting a major reworking of how git worktrees are used to take advantage of first-class support in Claude Code and Codex and other related removals/changes.
#15 𝕏
Santiago made TinyFish’s web search and fetch APIs free for all AI agents, giving you structured JSON search and clean URL fetch via a simple CLI.
#16 𝕏
Santiago warns that teams spending time “hardening” file-based implementations for concurrency and search are essentially rebuilding a database, so his benchmarks show files are fine for prototypes or single-user agents but databases are a must for multi-user, multi-agent pro...
#17 𝕏
Guillermo Rauch flags an open-source solution that scales via parallel cloud sandboxes, is battle-tested on massive repos, supports BYO agent/model/key, and offers an extremely easy-to-use CLI.
#18 𝕏
Peter Yang presents a 3-layer AI prompt—Functional specs, Visual wireframe, and synthetic JSON data—to generate polished app prototypes. By swapping only the JSON layer you can instantly pivot content (e.g., downtempo to psychedelic rock) without reworking the other layers.
#19 𝕏
Garry Tan warns “This is the worst the AI will ever be,” highlighting a Guardian report where AI outperformed doctors in Harvard’s emergency‐triage diagnoses trial.
#20 in
Michael Leibovich leveraged Claude Code (with Obsidian and Earmark) to compress 20 years of his leadership context into an always-on AI chief of staff, slashing meeting overload, context-switching and decision fatigue with personalized, non-generic outputs.
#21 in
Tal Raviv and Aman Khan host a live screenshare to build a company-wide AI OS from scratch, syncing high-quality context across teams. They’ll demo why local setups beat web, why cloud drives beat GitHub, and how to share agent skills.
#22 𝕏
Lenny Rachitsky posted the full YouTube conversation expanding on his AI agency clip, detailing his launch playbook, productized service structure, and tiered pricing model. He also shares early client ROI insights.
#23 𝕏
Lenny Rachitsky relays @mschoening’s insight that in today’s world skill gaps are no longer the main barrier—true progress hinges on giving people the agency to actually use those skills.
#24 𝕏
Peter Yang unveils a 3-layer prompt framework—functional spec, wireframe layout, and synthetic-data JSON—for generating polished AI prototypes (demonstrated with a music discovery app). By swapping only the JSON layer, you can instantly refocus your app’s content (e.g.