boldstart ventures✓

Clay @clay

Introducing auto-mode for Claygent. It picks the AI model for each row of research, balancing answer quality AND cost. Now on by default for new Claygent columns: - 47% cheaper than sending every task to the most capable model. - <1 second to choose the model for each research task. - You can still pick your favorite model anytime. Super cool thing: We post-trained an open-weight model with @appliedcompute to handle the routing. It runs on their AC2 platform and handles hundreds of millions of queries a month. @jeffbarg, Vyshnavi, @rhythmrg, and @_brylee10 break it down below.

Manya Ghobadi @systalyze

@systalyze is #1 in output speed on @ArtificialAnlys for GLM 5.2, in both reasoning and non-reasoning modes. → 538 tok/s: 1.7× faster than @baseten and @CoreWeave , 5.5× @Zai_org own endpoint → Cheapest provider on non-reasoning Try it:

Spectro Cloud @spectrocloudinc

We've been saying for a while that local inference can cut your devs' AI coding bill without slowing them down. This blog brings the receipts, from how we tuned an AMD Instinct Coder box to where it starts costing less than hosted APIs. Read it here 👉

MTS @MTSlive

Topos Bio CEO Ryan Zarcone explains why designing drugs for Parkinson's and Alzheimer's is almost impossible for humans, but a perfect problem for AI: "When a chemist designs a drug for a protein, the model is static. I have this protein, it has a shape, and I’m trying to fill that shape with a drug." "These really dynamic proteins, like those involved in Parkinson’s and Alzheimer’s, don’t have one shape. They have potentially thousands of different shapes that they’re constantly interconverting between, kind of like someone who’s dancing." "That becomes basically impossible when you tell a human, design something for these thousand things at the same time. It’s really only been possible to start doing that with AI." @ToposBio

ToposBio @ToposBio

Missed us on MTS? Here’s a clip from our conversation about modeling dynamic proteins and designing drugs for challenging disease targets. Thanks @MTSlive for having us!

MTS @MTSlive

Topos Bio CEO Ryan Zarcone explains why designing drugs for Parkinson's and Alzheimer's is almost impossible for humans, but a perfect problem for AI: "When a chemist designs a drug for a protein, the model is static. I have this protein, it has a shape, and I’m trying to fill that shape with a drug." "These really dynamic proteins, like those involved in Parkinson’s and Alzheimer’s, don’t have one shape. They have potentially thousands of different shapes that they’re constantly interconverting between, kind of like someone who’s dancing." "That becomes basically impossible when you tell a human, design something for these thousand things at the same time. It’s really only been possible to start doing that with AI." @ToposBio

ToposBio @ToposBio

Understanding protein dynamics can open new paths to treating difficult diseases. We spoke with GamesBeat about Topos-2, Topos-Bind, and our approach to designing drugs for dynamic proteins. Thanks @deantak for the conversation!

Generalist @GeneralistAI

Our CEO Pete Florence on The Generalist podcast, talking all things robots and how we're building for the physical world. @mariogabriele @thegeneralistco Listen to the whole conversation - link in the comments below

Madison Mills @MadisonMills22

SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here:

Microsoft Threat Intelligence @MsftSecIntel

The China state-aligned threat actor Storm-0940 used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. Microsoft Threat Intelligence assesses that Storm-0940 was responsible for the recently disclosed ConnectWise supply chain compromise of a multitenant application. Our investigation found that Storm-0940's use of ARTEX accelerated the speed of compromised application abuse to authenticate to target environments through applications that had already been granted access, allowing the threat actor to operate without requiring an interactive user sign-in. AI-accelerated follow-on activity included reconnaissance, application credential changes, and attempts to modify privileges that could enable expanded or persistent access. Read our blog for a detailed analysis and indicators of compromise (IOCs), along with Microsoft Defender detections, protection recommendations, and guidance on investigating service principal sign-ins and unexpected application credential or privilege changes.

Spectro Cloud @spectrocloudinc

As an NVIDIA DSX ecosystem partner, we integrated open-source NVIDIA Run:ai into PaletteAI. Together, we help customers put more of their GPU capacity to work without taking on greater operational complexity. How it works and how to try it 👉

Matthew Prince 🌥 @eastdakota

Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.

Snyk @snyksec

Finding vulnerabilities was never the hard part. Getting a fix your team will actually merge is. We built VulnBump to measure exactly that: real vulnerabilities from real projects, and a fix only counts if the build and tests still pass. LLMs alone: 59.6%. Evo Remediation Agent: 90%. Check the numbers at

Netpreme @netpreme

Sparse attention still needs the full KV cache. Our XM-Sparse Pilot VM testbed with SGLang HiSparse and Netpreme X-Mem™ shows lower TPOT and higher peak decode throughput vs. host DRAM for LLM inference. #LLMInference #SGLang

Ed Sim @edsim

fun panel yesterday to kick off @AISecSummit in SF, an unfiltered look at what's happening hacking is democratized, and only way for defenders to keep up is with more compute but so much has to be built to make it work some things at inception i'm thinking about - the physical frontier - who's securing the data centers and neoclouds, and the next headache will be billions of new intelligent devices coming online 👀 thanks @dee_bosa for moderating with @mnair1 @snyksec @JohnHultquist + Michael D'Angelo @OpenAI

Ed Sim @edsim

Celebrating a major milestone - 10 years of What's 🔥 - haven’t missed a week! 520 issues! What started out as notes to myself and a way for me to process the week that was, has turned into a labor of love since i do enjoy writing... 🙏🏻 for reading, sending notes, and being on this journey with me… For #520, back from a whirlwind week in SF and @AISecSummit with a few thoughts: → fight agent swarms with agent swarms—we don’t have the luxury of waiting → the agents can’t hold their own keys → AI is going physical…who secures the robots and drones? → who secures the neoclouds powering all of this? Plus a look at where I’m investing next 👀

a16z @a16z

Engineering leaders told Grepr founder and CEO Jad Naous they spend 5 to 20% of their infrastructure budget on observability, and it's buying less and less reliability. Jad, a former a16z partner, joins a16z's Martin Casado to discuss whether AI can run production systems as well as it writes code. Grepr sits in a company's log stream, can scan a petabyte of logs a day, cuts observability costs by about 90%, and is building AI agents that use those signals to troubleshoot production issues. How AI is changing engineering at Jad's 10-person company: - A data lake feature that once took one engineer about six months took Jad 10 days - The gap between okay and great engineers on his team "just multiplied 10X" - He runs 3-4 agent sessions at once, with a "tech lead" agent reviewing every PR Why running live systems is a harder problem for AI: - Jad: AI can almost tune databases, but can't yet keep servers running on its own - Martin: every bit of code doubles the ways a program can behave, which is why humans still own systems - The real world is messier than AI's training data, self-driving took 20 years and $170B 0:55 From Stanford to a16z to Grepr 7:00 Why observability spend stopped paying off 11:00 From autocomplete to AI that trains engineers 13:10 Why AI widens the engineer skill gap 14:55 Syntax vs semantics 18:00 Inside Jad's AI workflow 22:55 Six months of work in 10 days 24:50 Can AI reason about how systems really run? 30:10 Why AI SRE tools start in the wrong place 32:30 Can AI run a distributed cluster yet? 37:35 Why engineers buy extra capacity over predicting 39:15 How AI could trace failures in production 43:45 Why AI for ops is harder than for code 49:55 Can AI do a VC's job? 52:05 Why self-driving took 20 years and $170B 57:05 Customers skipped the AI pitch for the 90% savings YouTube: @jadtnaous @grepr_ai @martin_casado