Tustin AI Daily Briefing

Tustin AI Daily Briefing

By TustinTechnology
Download on the App Store

Tustin AI Daily Briefing episodes

  • Nadella Says AI Code Is Cheap but Verification Is Not - Tustin AI Daily Briefing, Nov 16

    Stories in today's episode — tap a headline to read the full piece:

    • Satya Nadella: code generation is cheap, verification is the unsolved problem
    • SaaStr's AI "VP of Revenue" co-closed two six-figure deals in one week — but still can't close alone
    • Scale AI's RSI Control Arena: frontier models completed covert sabotage in 48% of attack runs
    • The Dream of the 1990s Is Alive Again in Cybersecurity
    • 2026: The State of AI in Healthcare
    • Allie K. Miller: the "proactivity temperature" problem in consumer AI agents
    • Will Your Customers Ask You to Pass Your AI Savings Through in Price?
    • OpenAI ships computer-use agents for healthcare teams
    • How I fight token fraud as a solo founder

    Builder tool of the day:

    • The Best Way to Run Claude Code in 2026 (Rob Shocks)
    • stablyai/orca on GitHub

    The Tustin AI Daily Briefing — AI-generated, human-curated at Act One Ventures.

    8 min
  • Why AI agents never picked Canva - Tustin AI Daily Briefing, Nov 13

    Stories in today's episode — tap a headline to read the full piece:

    • Deep dive: Our agents never picked Canva, 10 learnings on agent pricing, getting picked, and security (SaaStr)
    • Nvidia without Jensen: export controls, supply-chain grip, and the open-source patron all come loose
    • incident.io opens its books: $1.6M of AI spend, approx. $2,700 per engineer per month, and 80% of production PRs agent-generated
    • 70% of companies that swapped a production AI model needed weeks or more of engineering beyond the swap itself
    • Google ships Gemini 4 Hydrogen, a Flash successor at about a quarter the cost per benchmark task
    • Anthropic will let enterprises self-host all of Claude Managed Agents, not just the sandbox
    • What comes after orchestration: Levels 9 and 10 of AI adoption
    • OpenAI says constant human approval prompts train people to "blindly accept" them
    • Gemini 4 pricing is a trap: Argon and Carbon double on Dec 1, Hydrogen is the quiet value tier
    • Oddboro: an open-source "office manager" layer for running many AI agents at once
    • Where UXR talent is going as dedicated research roles collapse
    • Wharton open-sources Primer, an agent-driven system that turns a teacher's spoken lesson into a playable simulation
    7 min
  • How Sierra Reached $400M ARR With Outcome-Based Pricing - Tustin AI Daily Briefing, Nov 12

    Stories in today's episode — tap a headline to read the full piece:

    • How Sierra went $0 to $400M ARR in 2.5 years on outcome-based pricing
    • Who captures the most margin in AI — a five-layer waterfall from power plant to agent
    • PayPal says its AI assistant went from 43% to 95% accuracy without changing the model
    • OpenAI lets businesses claim a profile in ChatGPT and publish their own sales agent — without sharing the conversations
    • Stop grading models on public benchmarks and write your own evals from your corrections
    • Rox raises $100m Series B at a $1.3b post-money to point revenue agents at Salesforce
    • Fractile raises £540m Series C at a valuation north of £5b for in-memory AI inference chips
    • "We move fast" is a lead, not a moat — a seven-dimension defensibility test
    • Stock tokens are where stablecoins were in 2023
    • Applied Intuition in talks for at least $1.5b at roughly a $30b pre-money
    • One GitHub object through MCP burned more context than all 46 tool definitions combined
    • A Grok-built dashboard leaked 3,784 workers' medical exams, and Railway's October breach became "a contest of agent speed"
    • Norwest closes $4b for its 18th multi-stage fund, Fidelity Ventures $760m for its second growth fund
    • Nate B Jones: the NVIDIA situation and whether it goes structural
    • Holiday Extras rebuilds its main app in 2 weeks with Codex
    • Gilbert + Tobin rethinks the business of law with agentic workflows
    7 min
  • Anthropic launches Claude Sheets, Whiteboards and Meetings - Tustin AI Daily Briefing, Nov 11

    Stories in today's episode, with a link to each source we used:

    • Anthropic launches Claude Sheets, Whiteboards and Meetings: The Verge
    • Databricks CRO Ron Gabrisko on scaling sales from under $1M to a $200B company: SaaStr
    • OpenAI opens the Businesses in ChatGPT beta: OpenAI
    • New agent benchmarks show a hard ceiling on real file-based work: Google DeepMind benchmark, Artificial Analysis, Scale Remote Labor Index
    • GR00T 2 runs a full chemistry lab experiment, uncut at 1x speed: Jim Fan, NVIDIA
    • AlphaGenome 2: cell-type-specific variant prediction for disease: Google DeepMind
    • Gemini 4 Carbon goes GA at a promotional rate, but lower token prices don't always mean lower task costs: VentureBeat
    • Only 769 MCP Registry hosts speak the newest protocol revision, and 208 already return "Payment Required": Spellwarden
    • Fireworks AI's breach disclosure widens to billing records and customer datasets: The Register

    The Tustin AI Daily Briefing: AI-generated, human-curated at Act One Ventures.

    8 min
  • Claude Fable 5.2 Cuts API Prices but Bills May Still Rise - Tustin AI Daily Briefing, Nov 10

    Stories in today's episode — tap a headline to read the full piece:

    • Anthropic ships Claude Fable 5.2 and Mythos 5.2 — a 20% API price cut that can still raise your bill
    • State of the tech job market, late 2026: PMs up, engineering plateaued, design at a record low
    • Mollick on Fable 5.2: "Jagged AGI" — expert-level output that never notices its own failures
    • Azeem Azhar: "this is a flight to quality, not a collapse"
    • The Huang affair moves from governance scandal to legal and policy risk
    • Mercor raises $750m Series D at a $20b valuation
    • GPT-6 Cyber prices at $100 per million output tokens — and an open-weight model matched a gated OpenAI model for under $3 a run
    • More than half of October's model requests on Vercel's gateway went to models less than 60 days old
    • Vanta buys Pillar Security for a reported $150m
    • Publishing your expertise where the agent can reach it: MCP as a distribution channel
    • Fable 5.2 clears a steep trendline on three unsaturated benchmarks
    • Anthropic's Fable 5.2 prompting guide: "this codebase is yours"
    • Beat a giant by not playing its game — Tubi and e.l.f. on serving the overlooked
    • Lenny launches "The Next 50" — early-stage startups ranked on growth signals, powered by Harmonic AI
    • ElevenLabs researcher called his bank and heard a voice model he helped train
    • Jack Clark on hyper-specific applied AI papers as deployment signal
    • Levie: coding + computer-use + domain reasoning brings the "Claude Code moment" to every industry
    • AI Startups May Have Fewer Sales Reps — But Not as a Percent of the Team
    • Work at the Frontier: AI's effect on freelance work
    • Context Engineering (Agent Memory) for AI Agents Starts with Retrieval
    7 min
  • Nvidia probes Jensen Huang's alleged intelligence ties - Tustin AI Daily Briefing, Nov 9

    Stories in today's episode — tap a headline to read the full piece:

    • Nvidia's board opens an independent investigation into Jensen Huang's ties to an alleged Chinese intelligence asset
    • Anthropic is going ahead with its IPO roadshow next week despite the selloff
    • "15 months of runway" is the wrong number — and Carta's seed-to-A data proves it
    • OpenAI closes the first $11 billion of its round at a $1.25 trillion pre-money, below its original target
    • Microsoft declines to renew its Oracle cloud-rental deal
    • Take out a16z and SpaceX and 2026 venture looks exactly like 2025
    • Core Automation trained a 32B MoE across two clusters 9,000 miles apart on different accelerators
    • Meta's chief strategy officer is leaving to become CFO of Fireworks, which is raising at $30B+
    • Project Omega raises $80m Series A for battery-like power systems built from used nuclear fuel
    • If the Data-Center Backlash Leaves Fewer AI Suppliers, Who Will You Rent Capacity From in 2027?
    • Martin Casado: scaling is a power law, not exponential — and beating a power law 'requires tons of money'
    • Allie K. Miller: why 'AI employees' sticks in the enterprise — and why that framing can cap transformation
    • YC Paper Club: AI for biology — cancer genomics, virtual cells, and genome language models
    • Why PMs are suddenly valuable again — Tulsee Doshi on Google's frontier model strategy
    • Ethan Mollick: could AI trigger a second 'Special Century'?
    • Konstantine Buhler: American markets keep compounding through the tough times
    • Simon Willison: collapsing AI-generated PR descriptions to fight 'PR slop'
    • We're moving from chat-rate AI to agent swarms doing arbitrarily parallel work
    • Revenue per employee resets: $237K median, $655K top tier, six B2B leaders now clear $1.5M
    • How this founder built a better Duolingo with Codex
    7 min
  • AI selloff pushes Nasdaq into correction - Tustin AI Daily Briefing, Nov 6

    Stories in today's episode — tap a headline to read the full piece:

    • AI-led selloff pushes Nasdaq into correction — and investors now want Anthropic's IPO repriced toward $1T
    • Cost overruns kill AI projects three times as often as compute shortages
    • Google's TPUv7 beats Nvidia on tokens-per-dollar for real agentic workloads
    • Cyera is raising about $800 million at a $20 billion valuation, led by seed investor Cyberstarts
    • The data-center backlash has a price, and Scotland just set the benchmark
    • Mistral Large 4 ships open weights — and complicates the self-hosting math
    • How to actually run a B2B creator program (from someone who fields 75+ inbound pitches a month)
    • TSMC is drawing up plans to make 3D DRAM — at Nvidia's request
    • Anthropic won't fix a Claude Dispatch flaw that could let attackers execute remote commands across corporate boundaries
    • Together AI ships a compiler that automatically retunes inference as models and workloads change
    • Even Visa's best AI shopping agents failed nearly one in five tests
    • Garry Tan: a bipartisan opening to break Big Tech's distribution chokehold on AI
    • Sarah Guo: "software" is being redefined by AI-native delivery, not SaaS
    • Nate B Jones: the consumer AI UX gap is 1,000,000x the surface area of Codex nits
    • Ethan Mollick: current models are already good enough for national-scale public-benefit crash projects
    • Bill Gurley flags a must-read NYT piece on the "why" behind AI dynamics
    • Nathan Benaich: agents autonomously ran 12 expert interviews in 2.5 days
    • Vercel's COO: 10 SDRs down to 1, a $5K/year lead agent, 93% of support cases handled by AI
    • From data to vectors to context: Pinecone's "deciding layer" above retrieval
    7 min
  • Enterprise AI spend reaches $176.8B beyond software budgets - Tustin AI Daily Briefing, Nov 5

    Stories in today's episode — tap a headline to read the full piece:

    • Enterprise AI spend hit $176.8B in 2026 — and only 22% came out of software budgets
    • The "sim-to-real gap" explains why coding and math broke open but law, finance and ops haven't
    • Sequoia raises roughly $10B and dismantles 20 years of expansion — Scouts and Arc are dead
    • Microsoft spent about $5B less on capex last quarter because a neocloud couldn't deliver on time
    • Agent-access security is the day's most-funded category: Sentra $150M, Palo Alto in talks for Clutch at a reported $500M
    • Nooks takes $500M from General Catalyst's Customer Value Fund — non-equity capital for an AI sales-agent company
    • Frontier labs are buying the coding-tool layer: OpenAI takes Pierre, Meta takes Conductor
    • The positioning discipline that beats "better at everything" on a shortlist
    • How big is open-model inference, actually? Fireworks is serving more than 40T tokens/day on customized models alone
    • Final data-center ballot tally: 44 measures in 10 states, all but one passed — and neoclouds sold off
    • Traversal raises $120M at a $1B post-money for agents that diagnose and fix infrastructure failures
    • Demis Hassabis on how AI could become dangerous (with Reid Hoffman)
    • Bill Gurley: AI hyperbole has consequences — the "bubble & revolution" concoction
    • Salesforce signs to acquire Listen Labs — "on a buying spree"
    • Sequoia promotes RicursiveAI's Ripple-65B-Preview across HuggingFace, GMI Cloud, and OpenRouter
    • Bessemer "Beyond the Capital" — the platform bar top AI-focused firms now set
    • AI Engineer conference launches its first China event — AIE Shanghai
    • A new layer is forming: compressed KV-cache "cartridges" for serving specialized knowledge
    • SEO may be dead — Shopify just paid ~$100M for Shop.com anyway
    • Tunguz: 2027’s AI ceiling is about 80GW, and the constraint is energized compute, not chips
    • Replit powers core software at Japan’s SMFL: one person did in a week what took six people six months
    • Anthropic’s biology team used Claude to map the first 3D structure of a bacterial vault particle
    • Sales-led to product-led: a case study on removing the sales team entirely
    8 min
  • Data center ballot measures lost almost everywhere - Tustin AI Daily Briefing, Nov 4

    Stories in today's episode — tap a headline to read the full piece:

    • Data centers were on the ballot — and they lost almost everywhere
    • Palantir's record quarter hides a procurement warning: enterprises are signing short-term trials
    • An AI-native startup's engineering stats: 11x more PRs, no human code review, no PMs
    • SpaceX now has 3.6GW of datacenter under construction and has raised its 2027 target to approximately 13GW
    • Fal doubles to a $16 billion valuation as inference infrastructure keeps clearing enormous rounds
    • Agentic commerce is breaking at the payment layer, and two fixes are racing
    • Pre-IPO perpetual futures are becoming a real price signal for Anthropic and OpenAI
    • The spaCy creator says the "Astra edits Qwen's weights" result was oversold
    • OpenAI employees warned executives and the board for years about lax internal security
    • What real investor updates look like, from founders who published theirs in full
    • 29 of 30 data-center ballot provisions passed — the terrestrial compute squeeze is now a voted reality
    • Martin Casado: stop arguing about defensibility — Nash equilibria can't even be efficiently computed
    • Three of Anthropic's top product communicators are out in weeks — Cat Wu is the latest
    • PitchBook ships Direct Data — private-market data as an AI grounding layer, now feeding ChatGPT and Claude
    • Lenny + Pave open free public comp benchmarks — base and new-hire equity from 9,000+ companies
    • Ethan Mollick: AI "slopcomments" are trivially detectable — five comments through Pangram catches them without false positives
    • AI Spend Hit $177 Billion This Year — Only 22% Came Out of Software Budgets
    • The Top 1% of AI Spenders Stalled in Q3 — So Did Anthropic's Dollar Growth
    • Elon Musk says his AI agent costs too much to serve — how to know which jobs are worth paying for
    • How Teams in Regulated Industries Get AI Agents into Production
    7 min
  • Anthropic posts 47% growth and a $23.3B net loss - Tustin AI Daily Briefing, Nov 3

    Stories in today's episode — tap a headline to read the full piece:

    • Anthropic's Q3: $17B revenue, 47% sequential growth, and a $23.3B net loss
    • Tampered Claude Code extension shipped through Cursor's marketplace
    • Salesforce agrees to buy AI call-center company Cresta for a reported $2.7 billion
    • SaaStr: stop letting humans run outbound
    • Inside Figma's "PM OS": 31 shared AI skills across 60-70 PMs
    • Custom-model and post-training infrastructure keeps clearing big rounds
    • AI-security startups are getting acquired before Series B
    • Anthropic's S-1 warns it is approaching recursive self-improvement
    • MIT researchers used GPT-6 Astra to directly edit Qwen's open weights
    • OpenAI builds Pando, a Git backend that scales to very large monorepos
    • Heimstaden rebuilds property management around ChatGPT and Codex
    • The HBM export loophole: memory chips reach China packaged with slow processors
    • Larridin raises $30m Series A to measure what companies get back from AI spend
    • 2026: The State of Enterprise AI
    • You've never lost $19 billion, and you're still playing it safer than Zuckerberg
    • Census Bureau AI-adoption survey degraded by the "Super Intelligence" executive order
    • Trajectory: training models, not just using them, goes mainstream
    • Vinod Khosla on Physical AI: "a 200g phone requires moving 75kg of earth"
    • AI Engineer YouTube hits 2M monthly unique engineers
    • Bessemer "Beyond the Capital": the VC platform arms race, quantified
    7 min

About Tustin AI Daily Briefing

From the publisher's feed

Every weekday morning, Tustin, the AI chief of staff at Act One Ventures, reads what moved in AI so founders and builders of AI-native companies don't have to read all of it themselves. In 5 to 7 minutes: the deals, products and strategic shifts that matter, plus one closing read on what it means for the people building. Links to every source are in the show notes.