DevPick

Top topics

Find news and summaries about the products and technologies most covered in public news.

News (page 124)

A tool using AI agents with LangGraph to execute, benchmark, and profile C++ CUDA kernels with nsight has been released.

  • •The creator utilized AI agents as a learning opportunity for LangGraph while optimizing CUDA kernels.
  • •The AI agent receives a C++ CUDA test harness to execute kernels and obtain benchmarks.
  • •The agent also features profiling capabilities using nsight.
Read the original ("Show HN: Agentic CUDA Kernel Optimizer")(opens in a new tab)#AI Agents#Generative AI Tools & Use#Semiconductors & Infrastructure
9/25/2026
AI summaryEstimated from title

Reports indicate Oracle has a contract structure requiring data center investors to pay even if sites are not receiving power.

Unread: Ink and Switch Interactive Homepage

Hacker NewsHN submitter: iFreilicht
Sign up to see

The interactive homepage of 'Ink and Switch,' an independent research lab exploring the future of tools for thinking, has been released.

  • •'Ink and Switch' is an independent research lab exploring the future of tools for thinking.
  • •The lab is researching and developing new interfaces and tools to support thinking processes.
  • •Specific product names and technical details cannot be confirmed from the provided information.
Read the original ("Ink and Switch Interactive Homepage")(opens in a new tab)#Research & Papers#Generative AI Tools & Use
9/25/2026

Venture capital firm Lightspeed revealed its plan to target a $250 million raise for a new India-focused fund, with a focus on early-stage AI companies.

  • •Lightspeed has set a $250 million fundraising target for its new India-focused fund.
  • •This is the first time the company's India fundraising cycle will be synchronized with its global fund.
  • •As part of a strategic shift to shorten investment periods, the firm is prioritizing investments in early-stage AI.

Unread: Jev Based Code Review

Hacker NewsHN submitter: namanbhulawat
Sign up to see

Jev, a review tool focusing on code behavior rather than just diffs, prioritizes human attention while OpenAI explains the changes.

  • •Jev focuses on actual code behavior rather than just diffs during reviews.
  • •The design prioritizes human attention, with OpenAI handling the explanation of changes.
  • •It is available as a local CLI, agent skill, and GitHub extension.
Read the original ("Jev Based Code Review")(opens in a new tab)#Generative AI Tools & Use
9/25/2026

An ML engineer analyzed how TypeSafe AI's non-autoregressive model Jev impacts classifiers in production environments.

  • •TypeSafe AI's Jev is positioned as a non-autoregressive System One model.
  • •The discussion covers its application scope in production classifiers and the relationship between calibration and accuracy.
  • •The author outlines plans for testing the model from an ML engineer's perspective.

Reports indicate that Microsoft has effectively ended the Copilot Plus PC brand, which it actively promoted about two and a half years ago.

  • •Microsoft initially strongly recommended Copilot Plus PCs as devices with built-in AI processing capabilities.
  • •According to reports, Microsoft and Qualcomm are in a situation where they acknowledge the brand is dead.
  • •The move to discontinue the brand became clear through reporting by Zac Bowden of Windows Central.

Waymo demonstrated rapid business expansion based on fleet data, including a 49% increase in vehicles in Texas.

  • •Waymo expanded its vehicle fleet size in Texas by 49% over the past month.
  • •There are other hotspots where the business is actively expanding, besides Texas.
  • •Fleet data shows that Waymo's autonomous driving service is scaling rapidly.

Startup Nexterity developed a robot that tightens or loosens four bolts at once, automating plumbing in a size that fits a Pelican case.

  • •Nexterity's robot can tighten or loosen four bolts at a time.
  • •The robot is compact enough to fit in a Pelican case.
  • •It aims to automate difficult and dangerous tasks in plumbing work.

A method was introduced where Opus 5.5 writes and renders explainer videos on OpenComputer's serverless agent simply by entering a URL or prompt.

  • •By entering only a URL or prompt, Opus 5.5 handles the video structure and rendering.
  • •The processing occurs on OpenComputer's serverless agent without using a dedicated video generation model.
  • •This approach enables content generation, such as launch videos, using only text-based AI models.
Read the original ("Opus 5.5 is good at explainer videos")(opens in a new tab)#Image / Video / Audio Generation#AI Agents
9/25/2026