boldstart ventures

Our XM-Sparse article explores how Netpreme X-Mem™ integrates with SGLang HiSparse to provide high-bandwidth access to offloaded KV caches. Read it here: netpreme.com/blog/xm-sparse-… Sparse attention is changing what AI systems need from memory. DeepSeek and Z.ai’s combined share of OpenRouter’s publicly reported tokens rose from 23% in July to 41% in October.* Meanwhile, research is advancing on several fronts: 🔹 Smarter selection of relevant context 🔹 Hierarchical search for efficient token retrieval 🔹 Reuse of KV data and token selections across layers 🔹 Hybrid architectures combining sparse retrieval with local or linear attention Agents accumulate long histories. Video creates enormous token sequences. Yet each decode step often needs only a small subset of that context. The hardware opportunity follows: retain KV history in a large warm memory tier, keep the active working set in HBM, and fetch selected data over a high-bandwidth fabric. At Netpreme, we’re building toward memory capacity that scales independently of compute, with scale-up bandwidth connecting GPUs to a shared warm memory tier. *The chart uses DeepSeek + Z.ai lab totals as a proxy for sparse-attention adoption, not a model-by-model count. Source: OpenRouter, via CrbonFree: crbonfree.com/case-studies/o…

Since January I have spent ~120 hours (three full work weeks, a lot of it while "on vacation") building a robust personal operating system. It focuses more on interlinkages vs. one-offs. I'd love to trade notes on architecture with like-minded folks. ellenchisa.substack.com/p/tr…

From a third of cloud spend toward 5 to 7%. Barry Carter, ex-7-Eleven CTO, on the observability math. Watch: youtu.be/gkPwxtZrrUw

Forrester says AI coding agents are “moving from developer assistants to active participants in development workflows.” Snyk has been named a Leader in The Forrester Wave™: Agentic Development Security Tools, Q4 2026, with the highest score in the Strategy category. We believe that reflects our focus on securing software where it is created, which now includes the agents creating it. Get the complimentary report: vist.ly/5rdgq

KPK’s agents acted within seconds of an alert, from Morpho exits to Gearbox pauses. This is what Hypernative is built for: giving teams the real-time signal to take whatever action they need, when it matters. We're very happy to be part of @kpk_io's stack and play our part in a positive result.🤝

THE BLOCK: Brazilian crypto exchange Mercado Bitcoin confirmed a security incident and temporarily suspended deposits and withdrawals, saying customers were unaffected. Blockchain security firm Hypernative earlier flagged more than $40 million in suspicious transfers across multiple networks, according to local outlet BitNotícias.

HIRING: FASTEST GROWING CYBERSECURITY STARTUPS (EARLY STAGE) Congrats to the 2026–2027 @FortuneMagazine Cyber60 winners! Here are the 20 early-stage companies on the list: A Security - autonomous offensive security that finds, proves, and fixes real attack paths @CylakeAI - on-prem security data platform for air-gapped and sovereign environments @straikerai - discovers, red-teams, and protects enterprise AI agents @P0Security - runtime access control for both humans and AI agents @air__security - a context firewall that vets the skills, MCPs, and plugins AI agents use @neo_ai_security - the control layer for AI agents running inside enterprise software Act - removes unnecessary cloud access paths before attackers can use them @glowio - AI-powered endpoint security built on prevention @artemis_sec - AI-native detection and response that adapts to each environment @scanner_dev - security data lake that searches every log in seconds @keycardai - identity and access for AI agents with task-scoped credentials @DropzoneAI - AI agents that investigate alerts and hunt threats for the SOC @jetstream_sec - governance and control layer for enterprise AI Bold Security - on-device AI that stops data loss and insider risk Astelia - exposure management that shows which vulnerabilities are actually reachable @ProphetSec - an agentic AI SOC that triages, investigates, and responds to alerts @CormaAI - foundation models built for cyber defense @Ent_Security - intent-aware security that stops risky actions by people and AI agents Frontier Security - autonomous defense for enterprise AI Oz Security - next-generation network security for distributed organizations Career pages for all 20 are in the comments 👇 @lightspeedvp @awscloud

Software Factories are extremely valuable, but hard to build. @tessl_io's new product may have just solved that 🙂 - Simple onboarding turns a repo into a factory in moments. - Template loops for code review, vulnerability remediation, cost optimization and more. - Built-in self-improvement for loops, compounding your factory’s value. - Interactive factory floor for tracking activity and solving problems. - Clear and simple metrics to know what you’re getting. Tessl factories are fully built on open skills, which you create, hone and own indefinitely. We help you create skills from your history, and the factory starts to self-improve them from here. Lastly, Tessl factories are declarative and easy to understand and manage. Config files composing skills and environments into actions, actions and triggers into loops, and multiple loops into a factory. This is the equivalent of Infrastructure-as-Code for factories. Instead of a bunch of disparate scripts you piece together to understand what’s happening, they codify what the factory does. This means Tessl factories aren’t just easy to setup, they’re easy to manage too. At the end of the day, our job as tool builders is to delight our users. Solve real problems with strong technology and great UX, that help them run fast. I feel the Tessl Factory is getting this job done. If it wasn’t obvious, I’m excited by our new product 🙂 At the end of the day, our job is delight our users, and I firmly feel the Tessl Factory does so. As part of the launch, we’re offering $200 free credits to all users. Try it out, and let us know what you think - I can’t wait to see what you build.

What's the future of GTM? We asked some of the top minds in the industry at SCULPT. BIG thank yous to @brendanjshort, Faris Sumadi, Erik Parry, Kyle Homovec, Johan van Tienen, Oriol Serra, Noah Burkholz, Sungwoo Chon!

The Binance breach is the largest crypto theft on record, with $6.3B drained from hot wallets in under an hour. The loss raises a critical question for any organization managing digital assets: what stops your systems from signing a malicious request? Our review covers what happened, how the attack worked, and what it means for teams across the industry. Read it here → buff.ly/PWrIbFV Dan Caspi walks through the attack live on Wednesday 4 November, 10am ET / 3pm GMT → buff.ly/B6YtezA

The 2026 American Dynamism 50 is here: our annual list of 50 companies, from within the a16z portfolio and beyond, building for America's future. This year's theme is Production is Deterrence. "America's ability to prevent conflict depends on what it can build, how fast it can deliver, and how long it can keep going." A new generation of defense companies is moving from prototypes to full-scale production, and the industrial base has to keep pace. This year's list is organized around three pillars: 1. Systems: the drones, ships, missiles, aircraft, and spacecraft that deter conflict. 2. Factories: the plants, tools, and software that build them. 3. Inputs: the materials and power they're made of and run on. Get all three right, and the impact reaches well beyond defense, to skilled jobs, revitalized communities, and the supply chains behind energy, housing, and everyday goods. Meet all 50: a16z.news/p/ad-50-2026-produ…

The Depository Trust & Clearing Corporation (@The_DTCC) Tokenization Service is live. Blockdaemon can provide firms with the infrastructure they need to connect on day one. blockdaemon.com/ecosystems#d…

Not all chips are fried. Read more about GEN-1.5 and sign up for the waitlist in the comments below.