Posts

Before: Multiple Simple Tools (Human-Centric Design)

Email Operations - Fragmented Approach:
┌─────────────────────────────────────────────────────────────┐
│                    Claude Decision Tree                      │
├─────────────────────────────────────────────────────────────┤
│  Email Task → Should I...?                                 │
│               ├─ draft_email.rb (2 params)                 │
│               ├─ send_email.rb (3 params)                  │
│               ├─ forward_email.rb (2 params)               │
│               ├─ find_and_draft_reply.rb (3 params)       │
│               ├─ read_email.rb (2 params)                  │
│               ├─ archive_emails.rb (1 param)              │
│               └─ safe_send_email.rb (4 params)             │
│                                                             │
│  Result: Decision paralysis, multiple API calls            │
│  Token Usage: ~150 tokens per operation                    │
│  Success Rate: 87% (context lost between calls)           │
└─────────────────────────────────────────────────────────────┘

                             ↓

CRM Operations - Scattered Functions:
┌─────────────────────────────────────────────────────────────┐
│  Company Research → Which tool?                            │
│                    ├─ find_attio_company.rb               │
│                    ├─ create_attio_company.rb             │
│                    ├─ update_attio_deal.rb                │
│                    ├─ enrich_company.rb                   │
│                    └─ validate_company.rb                 │
│                                                             │
│  Problem: Context switching between 5+ tools               │
│  Average: 3.2 tool calls per task                         │
└─────────────────────────────────────────────────────────────┘

After: Unified Complex Tools (AI-Centric Design)

Unified Email Tool - Comprehensive Context:
┌─────────────────────────────────────────────────────────────┐
│                unified_email_tool.rb                       │
├─────────────────────────────────────────────────────────────┤
│  Actions: [draft, send, reply, read, search, archive]      │
│                                                             │
│  Parameters:                                                │
│  ├─ --action (6 options)                                   │
│  ├─ --to, --cc, --bcc (recipients)                        │
│  ├─ --subject, --body, --body-stdin                       │
│  ├─ --from, --query, --folder (search)                    │
│  ├─ --sender, --in-reply-to (threading)                   │
│  ├─ --format [concise|detailed|ids_only]                  │
│  └─ --save-draft, --senders (batch ops)                   │
│                                                             │
│  Result: Single confident decision                          │
│  Token Usage: ~45 tokens (70% reduction)                  │
│  Success Rate: 94% (full context maintained)              │
└─────────────────────────────────────────────────────────────┘

                             ↓

Unified CRM Operations - Complete Workflow:
┌─────────────────────────────────────────────────────────────┐
│                attio_operations.rb                         │
├─────────────────────────────────────────────────────────────┤
│  Actions: [find, create, update, search, enrich,           │
│           validate_and_add, get_deals, update_deal]        │
│                                                             │
│  Intelligence Features:                                     │
│  ├─ Auto-enrichment (--enrich flag)                       │
│  ├─ Format optimization (concise/detailed/ids_only)       │
│  ├─ Source field management                               │
│  ├─ Deal pipeline integration                             │
│  └─ Notion page creation timing                           │
│                                                             │
│  Result: End-to-end workflow in single call               │
│  Average: 1.1 tool calls per task (89% reduction)         │
└─────────────────────────────────────────────────────────────┘

Performance Comparison

Metric Improvements:
┌─────────────────────┬─────────────┬─────────────┬─────────────┐
│ Measurement         │   Before    │    After    │ Improvement │
├─────────────────────┼─────────────┼─────────────┼─────────────┤
│ Token Usage         │ 150/op      │ 45/op       │ 70% ↓       │
│ Tool Calls          │ 3.2/task    │ 1.1/task    │ 66% ↓       │
│ Success Rate        │ 87%         │ 94%         │ 7% ↑        │
│ Decision Time       │ 2.3s        │ 0.8s        │ 65% ↓       │
│ Context Retention   │ Fragmented  │ Complete    │ Qualitative │
└─────────────────────┴─────────────┴─────────────┴─────────────┘

Implementation Principles:
1. Parameter-rich interfaces > Simple function calls
2. Comprehensive context > Modular operations
3. AI ergonomics ≠ Human ergonomics
4. Batch operations > Sequential calls

FFmpeg Analogy: Complex is Better for AI

Human Perspective (Intimidating):
┌─────────────────────────────────────────────────────────────┐
│ ffmpeg -i input.mp4 \                                       │
│   -vf "scale=1920:1080,fps=30" \                           │
│   -c:v libx264 -preset medium -crf 23 \                    │
│   -c:a aac -b:a 128k \                                     │
│   -movflags +faststart \                                   │
│   output.mp4                                               │
│                                                             │
│ Human Preference: Separate commands for each operation      │
│ ├─ scale_video.sh                                          │
│ ├─ set_framerate.sh                                        │
│ ├─ encode_video.sh                                         │
│ └─ optimize_audio.sh                                       │
└─────────────────────────────────────────────────────────────┘

                             ↓

AI Perspective (Perfect):
┌─────────────────────────────────────────────────────────────┐
│ Single command with complete transformation context         │
│                                                             │
│ Benefits for AI:                                           │
│ ├─ Understands full workflow intent                        │
│ ├─ Can optimize across all parameters                      │
│ ├─ No ambiguity about sequencing                          │
│ ├─ Complete error context                                  │
│ └─ Deterministic outcomes                                  │
│                                                             │
│ Result: AI confidently executes complex operations         │
└─────────────────────────────────────────────────────────────┘

These schematics illustrate the paradigm shift from human-centric (simple, modular) to AI-centric (complex, comprehensive) tool design that enables more efficient & confident AI operations.

Google commoditized its complements : free maps, free email, free browsers, a free mobile OS. They removed every toll booth between the user & search.

Anthropic’s strategy parallels Google’s, a natural extension of the strength of the core product, the model.

“BI became dashboards. And now it is re-expanding to business intelligence.”

— Colin Zima, CEO & co-founder, Omni

Colin describes how AI fuses structured & unstructured data, and why the future of business intelligence isn’t a better dashboard.

A support leader at one of Omni’s customers went through 75 pages of conversation with an AI to identify 10 categories of rep mistakes. The system read support logs, cited specific examples per rep, & suggested concrete changes. Not just a dashboard.

SpaceX announced a partnership with Cursor today : a $10 billion collaboration with a $60 billion acquisition option later this year.¹

The most important market in AI isn’t chatbots, search, or image generation. It’s coding. Cursor is the fastest-growing developer tool in history, at $2 billion in annualized revenue.²

To understand the deal, understand the stack. Winning in agentic coding requires three layers. Anthropic, OpenAI, & Google each own & operate compute, models, & distribution.

At the heart of every security team, there’s a database. That database records each time a user logs in, every packet of inbound traffic, & each attempted attack. Architected before AI, these SIEM systems are wooden shields in an era of autonomous attackers.

The consequences are mounting. Deepfake scams have stolen tens of millions. AI-generated phishing bypasses legacy filters. As Mythos has shown, the sophistication of attacks will only increase.

Shachar Hirshberg & Dan Shiebler saw this opportunity. Shachar led the Amazon GuardDuty product, scaling the business to over 80,000 customers. Dan built & led the 60-person AI/ML team at Abnormal Security. Together, they started Artemis to build a database to power defenses for modern security teams. Within a few months, they have more than a dozen production enterprise deployments & are processing over a billion events per hour. We are excited to partner with them at the Series A, along with our friends at Felicis, Brightmind, & First Round.

For the first time since the 2000s, technology companies are confronting the limits of their supply chain.

GPU rental prices for Nvidia’s Blackwell chips hit $4.08 per hour this week, up 48% from $2.75 just two months ago.¹ CoreWeave raised prices 20% & extended minimum contracts from one year to three.¹

“We’re making some very tough trades at the moment on things we’re not pursuing because we don’t have enough compute.” - Sarah Friar, OpenAI CFO¹

No one comes into a sales conversation without first asking an AI. The buyer journey has changed. Lena Waters, marketing leader behind DocuSign’s IPO, Grammarly & Notion, joined me on Office Hours to discuss what this means for your go-to-market.

The first phase of AI transformation is debt repayment. Most companies are agentically connecting go-to-market processes that should have been fixed years ago.

“Removing human coordination overhead and calling it transformation? That’s debt repayment. It’s real value, but it’s not a new paradigm.”

The demand for software is infinite. Kyle Daigle, GitHub’s COO, made the case concrete :

There were 1 billion commits in 2025. Now, it’s 275 million per week, on pace for 14 billion this year if growth remains linear (spoiler : it won’t.) GitHub Actions has grown from 500M minutes/week in 2023 to 1B minutes/week in 2025, and now 2.1B minutes so far this week.

But that’s not true for all roles. I use a 2x2 matrix that separates work along two axes : the ceiling of demand & whether the loop can be closed.

A researcher at Anthropic found out about a successful exploit when the model sent him an email. He was eating a sandwich on a bench outside.

Anthropic released Claude Mythos yesterday. Beyond the engineer’s lunch, the model has the potential to eat software’s.

In testing, Mythos found a 27-year-old bug in one of the most secure operating systems ever built, & a 16-year-old vulnerability in video software that conventional tools had examined five million times. Mythos is Anthropic’s largest model, roughly 10 trillion parameters, six times the size of any previous frontier model.

Two years ago, the idea of useful AI on your phone was fantastical. Siri couldn’t finish a sentence. Local models hallucinated nonsense.

Last week, Google released Gemma 4 E4B¹, a free model that matches GPT-4o and runs entirely on your phone.²

The next few weeks promise even more advanced pocket models. The market expects new releases from DeepSeek³, Qwen⁴, Kimi⁵ & Minimax⁶.

Frontier models don’t stay frontier for long. Within three to four months, you can run a model with similar performance on your laptop; 23 months later, you can run the same model on your phone.