Kevin Mandia (Mandiant founder, $5.4B exit) raised $255.5M Series B for Armadin at $2.5B valuation. Total: $445M+ in 9 months.
The playbook:
• Wedge: Traditional pen testing is slow, misses attack chains
• Solution: AI agent swarms test continuously, find vulnerabilities before attackers
• Proof: Series A showed PMF
• Velocity: Series B in 6 months (A16Z, Accel, GV, In-Q-Tel)
• Timing: AI agents now capable enough to replace human workflows
Proven founders don't start from zero—they see where old tools break and build the new category.
https://techcrunch.com/2026/10/01/kevin-mandias-new-agent-swarm-security-startup-armadin-raises-255-5m-at-2-5b-valuation/
Brewed by Startup Beaker.
The playbook:
• Wedge: Traditional pen testing is slow, misses attack chains
• Solution: AI agent swarms test continuously, find vulnerabilities before attackers
• Proof: Series A showed PMF
• Velocity: Series B in 6 months (A16Z, Accel, GV, In-Q-Tel)
• Timing: AI agents now capable enough to replace human workflows
Proven founders don't start from zero—they see where old tools break and build the new category.
https://techcrunch.com/2026/10/01/kevin-mandias-new-agent-swarm-security-startup-armadin-raises-255-5m-at-2-5b-valuation/
Brewed by Startup Beaker.
TechCrunch
Kevin Mandia's new 'agent swarm' security startup Armadin raises $255.5M at $2.5B valuation | TechCrunch
Kevin Mandia, best known as the founder of Mandiant, has a new startup that is using agent swarms to test and protect enterprises.
Founder Pedigree + AI Timing = $2.5B in 7 Months
Kevin Mandia's Armadin raised $255.5M Series B at $2.5B+ valuation, 7 months from stealth.
What: AI agent swarms run continuous offensive security tests, chaining vulnerabilities like real attackers.
Why it matters:
• Mandia's Mandiant sold to Google for $5.4B
• Fortune 500 + gov customers already in production
• a16z + Accel co-led; all prior backers returned
• $445M total raised in under a year
The play: Founder credibility + category timing compressed the path to unicorn. Watch the revenue metrics—this is priced on track record, not unit economics yet.
Source: TechCrunch, Armadin press release
#startups #AI #cybersecurity
Kevin Mandia's Armadin raised $255.5M Series B at $2.5B+ valuation, 7 months from stealth.
What: AI agent swarms run continuous offensive security tests, chaining vulnerabilities like real attackers.
Why it matters:
• Mandia's Mandiant sold to Google for $5.4B
• Fortune 500 + gov customers already in production
• a16z + Accel co-led; all prior backers returned
• $445M total raised in under a year
The play: Founder credibility + category timing compressed the path to unicorn. Watch the revenue metrics—this is priced on track record, not unit economics yet.
Source: TechCrunch, Armadin press release
#startups #AI #cybersecurity
125B AI Model on Your Gaming PC—No Cloud Required
Strata: run Qwen3.8-Flash-Next (125B) locally on RTX 20+ or AMD Radeon. 12 GB VRAM, 32 GB RAM, ~80 GB disk.
Speed: 94 tokens/sec on RTX 5070. Handles 262K context, reads images. OpenAI-compatible API.
How: Qwen released the model in August. Community dev (Niko1221) built Strata for consumer GPU inference. 2-3 bit quantization + one-click setup via Claude Code/Cursor. 10K+ GitHub stars in 48h.
Catch: Quantization trades precision for speed. 2-bit is rough; 4-bit on rented GPUs safer for production.
Takeaway: Local inference is now accessible. Test before scaling to cloud APIs.
https://github.com/Niko1221/Strata
Brewed by Startup Beaker.
Strata: run Qwen3.8-Flash-Next (125B) locally on RTX 20+ or AMD Radeon. 12 GB VRAM, 32 GB RAM, ~80 GB disk.
Speed: 94 tokens/sec on RTX 5070. Handles 262K context, reads images. OpenAI-compatible API.
How: Qwen released the model in August. Community dev (Niko1221) built Strata for consumer GPU inference. 2-3 bit quantization + one-click setup via Claude Code/Cursor. 10K+ GitHub stars in 48h.
Catch: Quantization trades precision for speed. 2-bit is rough; 4-bit on rented GPUs safer for production.
Takeaway: Local inference is now accessible. Test before scaling to cloud APIs.
https://github.com/Niko1221/Strata
Brewed by Startup Beaker.
GitHub
GitHub - Niko1221/Strata: Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference…
Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input. - Niko1221/Strata
19-year-old Zain Javaid raised $11M from a16z for Ghost—a $3,499 personal AI computer called Core.
The thesis: AI agents need local, sovereign hardware. Not cloud. Not corporate infrastructure.
Core specs:
- Nvidia RTX Pro 4000 Blackwell GPU
- Runs open-source models locally (Qwen, Gemma)
- Full encryption; Ghost can't access your data
- Accessed via phone app with voice mode
- First batch sold out in hours
Why it matters: Founders building privacy-first AI infrastructure have a massive wedge. The cloud era of AI is ending.
https://techcrunch.com/2026/10/05/at-19-ghost-founder-raises-11-million-to-build-a-3499-computer-for-your-personal-ai/
#AI #PersonalAI
The thesis: AI agents need local, sovereign hardware. Not cloud. Not corporate infrastructure.
Core specs:
- Nvidia RTX Pro 4000 Blackwell GPU
- Runs open-source models locally (Qwen, Gemma)
- Full encryption; Ghost can't access your data
- Accessed via phone app with voice mode
- First batch sold out in hours
Why it matters: Founders building privacy-first AI infrastructure have a massive wedge. The cloud era of AI is ending.
https://techcrunch.com/2026/10/05/at-19-ghost-founder-raises-11-million-to-build-a-3499-computer-for-your-personal-ai/
#AI #PersonalAI
TechCrunch
At 19, founder raises $11M for Ghost, maker of a $3,499 computer for personal AI | TechCrunch
Ghost's first product is Core, a personal computer designed specifically for AI agents that can take actions on a person's behalf.
🎇 BEYOND Accelerator is an acceleration program in Tashkent for startups that already have a product, revenue and ambition, and have outgrown their usual limits.
During the program, you will:
— strengthen your product and processes with AI;
— prepare for your round, from the financial model to the Data Room;
— connect directly with venture funds and angel investors;
— get one-on-one mentorship from experienced trackers;
— reach corporate partners and new markets.
Lead tracker: Bota Mussa, Founder & CEO of Bilimber, alumna of 500 Global, Antler and EF, with 300+ projects and 10,000+ hours of mentoring.
🎯 Who it's for: Pre-Seed / Seed startups with a working MVP and proven traction
📍 Format: in person, Tashkent
🗣 Language: RU / EN
20 October: application deadline
23 October: program kick-off
👉 Apply here
This project has been funded by UK International Development from the UK government and Government of Switzerland, together with the United Ventures.
During the program, you will:
— strengthen your product and processes with AI;
— prepare for your round, from the financial model to the Data Room;
— connect directly with venture funds and angel investors;
— get one-on-one mentorship from experienced trackers;
— reach corporate partners and new markets.
Lead tracker: Bota Mussa, Founder & CEO of Bilimber, alumna of 500 Global, Antler and EF, with 300+ projects and 10,000+ hours of mentoring.
🎯 Who it's for: Pre-Seed / Seed startups with a working MVP and proven traction
📍 Format: in person, Tashkent
🗣 Language: RU / EN
20 October: application deadline
23 October: program kick-off
👉 Apply here
This project has been funded by UK International Development from the UK government and Government of Switzerland, together with the United Ventures.
Claude Haiku 5.5 just matched GPT-6 Luna's price: $0.10/$0.50 per million tokens.
What happened: Anthropic cut Haiku's cost 10x in one year. Last October, Haiku 4.5 cost $1/$5. Now Haiku 5.5 is 10x cheaper and faster.
Why it matters for your stack:
• Haiku 5.5 now costs the same as OpenAI's cheapest model—no reason to overpay
• 1M token context window: handle full codebases, long docs, customer transcripts in one call
• Adaptive thinking: you control reasoning effort (low/medium/high) to trade speed for accuracy
• Multimodal: images, PDFs, screenshots—no separate vision API needed
The move: benchmark your top 3 tasks against Haiku 5.5 this week. If you're still on Haiku 4.5 or paying for premium models on simple work, you're leaving 80% cost savings on the table.
Source: Anthropic, Oct 7, 2026.
Brewed by Startup Beaker.
#AI #Startups
What happened: Anthropic cut Haiku's cost 10x in one year. Last October, Haiku 4.5 cost $1/$5. Now Haiku 5.5 is 10x cheaper and faster.
Why it matters for your stack:
• Haiku 5.5 now costs the same as OpenAI's cheapest model—no reason to overpay
• 1M token context window: handle full codebases, long docs, customer transcripts in one call
• Adaptive thinking: you control reasoning effort (low/medium/high) to trade speed for accuracy
• Multimodal: images, PDFs, screenshots—no separate vision API needed
The move: benchmark your top 3 tasks against Haiku 5.5 this week. If you're still on Haiku 4.5 or paying for premium models on simple work, you're leaving 80% cost savings on the table.
Source: Anthropic, Oct 7, 2026.
Brewed by Startup Beaker.
#AI #Startups
👍1
AGENT-TO-AGENT PAYMENTS: THE NEXT FRONTIER
Causa Prima raised $20M seed (Base10 led) to let AI agents from buyer and supplier companies negotiate payments directly. $98T in annual B2B payments stuck in manual approvals taking 8+ days.
How it happened:
- Founders met in Madrid pitching earlier startups; VCs passed but introduced them
- Years of friendship → Gebbing stepped back from Finoa; sketched agentic AI ideas with Ament
- Realized: internal AI agents wait days for human email replies
- Founded Causa Prima early 2026; brought in Stanislaus (ex-Centrifuge CTO)
- 25 employees, Madrid + Munich; expanding to NYC
Why it matters: Agent-to-agent networks are coming. First to own a vertical wins distribution and data moat.
https://fortune.com/2026/10/13/startup-causa-prima-20-million-billing-disputes-ai-agents/
#AIAgents #Fintech
Brewed by Startup Beaker.
Causa Prima raised $20M seed (Base10 led) to let AI agents from buyer and supplier companies negotiate payments directly. $98T in annual B2B payments stuck in manual approvals taking 8+ days.
How it happened:
- Founders met in Madrid pitching earlier startups; VCs passed but introduced them
- Years of friendship → Gebbing stepped back from Finoa; sketched agentic AI ideas with Ament
- Realized: internal AI agents wait days for human email replies
- Founded Causa Prima early 2026; brought in Stanislaus (ex-Centrifuge CTO)
- 25 employees, Madrid + Munich; expanding to NYC
Why it matters: Agent-to-agent networks are coming. First to own a vertical wins distribution and data moat.
https://fortune.com/2026/10/13/startup-causa-prima-20-million-billing-disputes-ai-agents/
#AIAgents #Fintech
Brewed by Startup Beaker.
Fortune
Startup Causa Prima raises $20 million to tackle billing disputes with AI agents | Fortune
Backed by Base10, Causa Prima’s software lets buyers’ and suppliers’ AI agents negotiate payment terms.
Boston Dynamics + Google DeepMind: Robots Get Reasoning.
Atlas humanoid robots now run on Gemini Robotics foundation models. Combines Boston Dynamics' hardware with DeepMind's multi-modal AI for perception, reasoning, and tool use in industrial settings.
Why it matters:
• Robots move from scripted tasks to adaptive reasoning—can now interpret environments and make local decisions.
• First wave: automotive manufacturing. Spot (robot dog) already deployed with AIVI-Learning for visual inspection.
• Opens B2B software layer: simulation data, safety tooling, audit logs, exception handling, and compliance workflows become defensible.
How it happened:
• Boston Dynamics spent 15+ years perfecting hardware dynamics and balance—the "athletic intelligence."
• Google DeepMind built Gemini Robotics (visual-language models for embodied reasoning) over 2+ years of research.
• Partnership announced CES 2026 (January); joint research kicked off immediately at both companies.
• First practical deployment: Spot + Gemini Robotics ER 1.6 for industrial inspection tasks (live now).
• Next phase: Atlas humanoid robots handling complex multi-step tasks in factories by Q3 2026.
Founder takeaway: If you're building industrial software, robotics tooling, or manufacturing ops—this is your wedge. The winners won't be robot makers; they'll be the teams that make robot actions reliable, auditable, and governable. Build the compliance layer, not the robot.
Brewed by Startup Beaker.
Atlas humanoid robots now run on Gemini Robotics foundation models. Combines Boston Dynamics' hardware with DeepMind's multi-modal AI for perception, reasoning, and tool use in industrial settings.
Why it matters:
• Robots move from scripted tasks to adaptive reasoning—can now interpret environments and make local decisions.
• First wave: automotive manufacturing. Spot (robot dog) already deployed with AIVI-Learning for visual inspection.
• Opens B2B software layer: simulation data, safety tooling, audit logs, exception handling, and compliance workflows become defensible.
How it happened:
• Boston Dynamics spent 15+ years perfecting hardware dynamics and balance—the "athletic intelligence."
• Google DeepMind built Gemini Robotics (visual-language models for embodied reasoning) over 2+ years of research.
• Partnership announced CES 2026 (January); joint research kicked off immediately at both companies.
• First practical deployment: Spot + Gemini Robotics ER 1.6 for industrial inspection tasks (live now).
• Next phase: Atlas humanoid robots handling complex multi-step tasks in factories by Q3 2026.
Founder takeaway: If you're building industrial software, robotics tooling, or manufacturing ops—this is your wedge. The winners won't be robot makers; they'll be the teams that make robot actions reliable, auditable, and governable. Build the compliance layer, not the robot.
Brewed by Startup Beaker.
Optical Switching for AI Just Scaled
Salience Labs raised $125M Series B for all-optical circuit switching in data centers. Light carries the signal—no electrical conversion.
Key metrics:
• 64-port all-optical switch demonstrated
• $8B market forecast by 2030
• Bullhound led; AMD Ventures + Intel Capital in
How they won:
- Founded 2021 by McKinsey operator + Oxford photonics researcher
- 10+ years of university research backing
- Validated 32-port prototype before scaling
- Showed live 64-port demo → oversubscribed round
Lesson: Deep science + credible proof = patient capital. Founders who ship silicon before pitching hype raise bigger rounds.
https://techfundingnews.com/oxford-photonics-startup-salience-labs-raises-125m-to-link-thousands-of-gpus-with-light/
#DeepTech #Funding
Salience Labs raised $125M Series B for all-optical circuit switching in data centers. Light carries the signal—no electrical conversion.
Key metrics:
• 64-port all-optical switch demonstrated
• $8B market forecast by 2030
• Bullhound led; AMD Ventures + Intel Capital in
How they won:
- Founded 2021 by McKinsey operator + Oxford photonics researcher
- 10+ years of university research backing
- Validated 32-port prototype before scaling
- Showed live 64-port demo → oversubscribed round
Lesson: Deep science + credible proof = patient capital. Founders who ship silicon before pitching hype raise bigger rounds.
https://techfundingnews.com/oxford-photonics-startup-salience-labs-raises-125m-to-link-thousands-of-gpus-with-light/
#DeepTech #Funding
Tech Funding News
Oxford photonics startup Salience Labs raises $125M to link thousands of GPUs with light
Salience Labs, a photonic chip startup based in Milton, Oxfordshire, has closed an oversubscribed $125 million Series B led by Bullhound Capital.
Databricks Just Raised $5B—Here's Why It Matters
Databricks closed a $5B Series I at a $43B valuation in Q3 2026, tying for the largest round of the quarter. This isn't just a funding milestone—it's a signal about where enterprise AI infrastructure is heading.
The wedge: Started as an open-source Apache Spark project (2013), then built a managed platform.
How they scaled:
- Wedge: Free, beloved OSS to enterprise adoption
- Distribution: Became the data layer for AI/ML teams
- Proof: 10,000+ customers, including 50% of Fortune 500
- Monetization: $1B+ ARR (per reports)
Why founders should care: Infrastructure plays that own the data layer win. Databricks proved you can build a $43B company by solving a real problem for data teams, not chasing hype.
Source: Crunchbase Q3 2026 Report
#startups #AI
Databricks closed a $5B Series I at a $43B valuation in Q3 2026, tying for the largest round of the quarter. This isn't just a funding milestone—it's a signal about where enterprise AI infrastructure is heading.
The wedge: Started as an open-source Apache Spark project (2013), then built a managed platform.
How they scaled:
- Wedge: Free, beloved OSS to enterprise adoption
- Distribution: Became the data layer for AI/ML teams
- Proof: 10,000+ customers, including 50% of Fortune 500
- Monetization: $1B+ ARR (per reports)
Why founders should care: Infrastructure plays that own the data layer win. Databricks proved you can build a $43B company by solving a real problem for data teams, not chasing hype.
Source: Crunchbase Q3 2026 Report
#startups #AI
❤2
Solo Founders Are Winning (And You Don't Need a Co-Founder to Start)
YC just analyzed 18 months of data. The stat: 19% of accepted companies now have solo founders—up from 5% a year ago. That's a 4x jump.
Why it's happening:
• AI coding agents handle what used to require a full engineering team
• Knowing what to build > knowing how to build it (the hard part)
• Experienced founders (late 30s, 40s, 50s) are having a resurgence—they have taste and know the dragons
• You can now spin up 20 coding sessions in a day to reach product maturity fast
The playbook:
1. Start solo if you have domain expertise or a clear problem
2. Add co-founders after you hit traction (not before)
3. Use AI agents to compress the "build" phase from months to weeks
4. Focus on knowing what to build—that's the bottleneck now
Median YC company now hits $20K MRR by demo day (vs $8K a year ago). Solo founders are part of that acceleration.
Source: YC's State of Startups 2026
YC just analyzed 18 months of data. The stat: 19% of accepted companies now have solo founders—up from 5% a year ago. That's a 4x jump.
Why it's happening:
• AI coding agents handle what used to require a full engineering team
• Knowing what to build > knowing how to build it (the hard part)
• Experienced founders (late 30s, 40s, 50s) are having a resurgence—they have taste and know the dragons
• You can now spin up 20 coding sessions in a day to reach product maturity fast
The playbook:
1. Start solo if you have domain expertise or a clear problem
2. Add co-founders after you hit traction (not before)
3. Use AI agents to compress the "build" phase from months to weeks
4. Focus on knowing what to build—that's the bottleneck now
Median YC company now hits $20K MRR by demo day (vs $8K a year ago). Solo founders are part of that acceleration.
Source: YC's State of Startups 2026
Modal just hit $1B ARR and raised $750M Series D at $15.75B valuation.
What they do: AI inference infrastructure—the picks-and-shovels layer for AI apps.
How they got here:
• Started with a wedge: serverless compute for ML engineers (2021)
• Shipped fast: went from $0 to $1B ARR in ~5 years
• Rode the wave: Series C ($355M, May 2026) → Series D ($750M, Oct 2026) in 5 months
• Valuation tripled in 4 months (May: $4.65B → Oct: $15.75B)
Why it matters: Infrastructure plays win in AI booms. If you're building AI apps, you need to own your distribution and unit economics—not just ride someone else's platform.
Brewed by Startup Beaker.
#AIinfrastructure #Funding
What they do: AI inference infrastructure—the picks-and-shovels layer for AI apps.
How they got here:
• Started with a wedge: serverless compute for ML engineers (2021)
• Shipped fast: went from $0 to $1B ARR in ~5 years
• Rode the wave: Series C ($355M, May 2026) → Series D ($750M, Oct 2026) in 5 months
• Valuation tripled in 4 months (May: $4.65B → Oct: $15.75B)
Why it matters: Infrastructure plays win in AI booms. If you're building AI apps, you need to own your distribution and unit economics—not just ride someone else's platform.
Brewed by Startup Beaker.
#AIinfrastructure #Funding
Agents Don't Need Apps—They Need Messaging.
Photon just raised $4.5M to let developers deploy AI agents directly into iMessage, WhatsApp, Telegram, and SMS. No app store. No download friction.
Why this matters:
• 40,000+ developer sign-ups in <1 year
• 10x revenue growth in 4 months (May–Aug)
• <3% churn, 5x message volume growth month-over-month
• 1M+ end users already reached
How they did it:
1. Founder built a bot texting friends as him over iMessage
2. Open-sourced it on GitHub → thousands adopted it
3. Realized: users already live in messaging apps
4. Built SDK + hosted infrastructure (99.95% uptime, HIPAA-compliant)
5. Went after regulated verticals (fintech, healthcare)
Key insight: 98% of Photon's usage is still open-source. The hosted tier captures enterprises needing compliance + reliability.
For founders: Distribution beats features. If your users live in messaging, meet them there—don't force them to download.
Brewed by Startup Beaker.
Photon just raised $4.5M to let developers deploy AI agents directly into iMessage, WhatsApp, Telegram, and SMS. No app store. No download friction.
Why this matters:
• 40,000+ developer sign-ups in <1 year
• 10x revenue growth in 4 months (May–Aug)
• <3% churn, 5x message volume growth month-over-month
• 1M+ end users already reached
How they did it:
1. Founder built a bot texting friends as him over iMessage
2. Open-sourced it on GitHub → thousands adopted it
3. Realized: users already live in messaging apps
4. Built SDK + hosted infrastructure (99.95% uptime, HIPAA-compliant)
5. Went after regulated verticals (fintech, healthcare)
Key insight: 98% of Photon's usage is still open-source. The hosted tier captures enterprises needing compliance + reliability.
For founders: Distribution beats features. If your users live in messaging, meet them there—don't force them to download.
Brewed by Startup Beaker.
👍2
🔐 Infisical: $40M Series B to Secure AI Agents
Infisical raised $40M Series B at $600M valuation, led by Elad Gil (doubling down from Series A). The secrets management platform is now cash-flow positive with 1,300+ customers including LG, Wise, and the US Government.
Why now? Three shifts:
• AI agents writing 10x more code = 10x more infrastructure to secure
• Attackers exploiting credentials as the entry point into systems
• Agents accessing production without human oversight
How they got here:
1. Built product developers love (30K GitHub stars, open-source moat)
2. Stayed cash-flow positive—didn't need to raise, chose to
3. Focused on developer experience + security (not a tradeoff)
4. Expanded from secrets → certificates → privileged access → Agent Vault
Key metric: 1,300 organizations trust them; hiring 100+ roles.
Takeaway: Founders who stay profitable longer have more leverage. Infisical raised because they could move faster, not because they had to.
Infisical raised $40M Series B at $600M valuation, led by Elad Gil (doubling down from Series A). The secrets management platform is now cash-flow positive with 1,300+ customers including LG, Wise, and the US Government.
Why now? Three shifts:
• AI agents writing 10x more code = 10x more infrastructure to secure
• Attackers exploiting credentials as the entry point into systems
• Agents accessing production without human oversight
How they got here:
1. Built product developers love (30K GitHub stars, open-source moat)
2. Stayed cash-flow positive—didn't need to raise, chose to
3. Focused on developer experience + security (not a tradeoff)
4. Expanded from secrets → certificates → privileged access → Agent Vault
Key metric: 1,300 organizations trust them; hiring 100+ roles.
Takeaway: Founders who stay profitable longer have more leverage. Infisical raised because they could move faster, not because they had to.