Matthew Turley | Fractional CTO

2.6K posts

Matthew Turley | Fractional CTO banner
Matthew Turley | Fractional CTO

Matthew Turley | Fractional CTO

@matthewrturley

Fractional CTO. I build MVPs and unbreak AI-built apps for SaaS founders. American in Paris, 3 kids. Need your backlog shipped?

Katılım Ekim 2009
1.1K Takip Edilen592 Takipçiler
Sabitlenmiş Tweet
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
Found a 19x sales lever hiding in a client’s CRM. Most SMBs don’t need “AI strategy” first. They need someone to look at their actual sales data. I analyzed ~4,000 leads for a specialty home improvement company and found: - leads with measurements closed 100x better - leads with photos closed 5.5x better - template-only follow-up: 1.08% close rate - template + personal reply: 20.29% close rate That last one is the money. The automation was working. The human follow-up system wasn’t. AI is most useful when it finds the boring operational leaks costing real revenue.
English
0
0
0
44
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
My work has basically evolved like this: → HTML/CSS (~2004) → static websites → PHP websites → WordPress themes → WordPress plugins → WordPress-powered applications → custom web apps → SaaS products → AI-assisted SaaS development → AI-native products → agentic workflows → AI agent orchestration layers → agent harnesses → autonomous build systems (today) Looking back, the progression was never really about the technologies. Each one was just a new abstraction layer. HTML and CSS taught me structure. PHP taught me logic. WordPress taught me leverage. Plugins taught me extensibility. Platforms taught me systems thinking. SaaS taught me product thinking. AI-assisted development taught me speed. Agent orchestration is teaching me coordination. The work keeps moving upward. From pixels, to pages, to workflows, to products, to systems, to systems that can create other systems. The deeper pattern is that I keep building closer and closer to the bottleneck. And right now, the bottleneck is no longer whether I can build the thing. It’s whether I can design, direct, verify, and control the machinery that builds it.
English
0
0
1
35
Minh-Phuc Tran
Minh-Phuc Tran@phuctm97·
❌ Complete a task ❌ Build an app to complete the task ❌ Prompt an agent to build the app to complete the task ✅ Build an agent harness to create the agent to build the app to complete the task
English
19
3
49
3.8K
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
@stevelauda_ You sure do, I should have checked. Fingers got ahead of my brain. Sounds like you are doing it right. Much better than throwing juniors into the mix expecting them to just figure it out (I worked at an agency like this before).
English
0
0
0
20
Steve Lauda
Steve Lauda@stevelauda_·
@matthewrturley I have a team hehe and I've done my best to create a leveling up system to make sure everyone to be on the same level
English
1
0
1
44
Steve Lauda
Steve Lauda@stevelauda_·
It's crazy to see a website build being charged above 5k USD by a well-known so-called "creative agency" but the quality looks like being done by a junior designer who are just getting started to freelancing.
English
24
1
87
6.9K
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
Can’t believe I waited this long. Finally made an iOS app to replace Telegram chat with my agents. Full control over the UI, cleaner context switching, better notifications, and a workflow that actually matches how I use them every day. Telegram was fine for proving the loop worked. But once agents become part of your daily operating system, the interface really matters.
English
0
0
0
42
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
@stevelauda_ Yeah, agree with you here. But you're solo right (I am)? I've usually seen this happen when they try to scale. Quality starts to slip as they are focused on other dynamics/metrics.
English
1
0
1
42
Steve Lauda
Steve Lauda@stevelauda_·
@matthewrturley for me, I always put myself in the shoes of my client. And also, I will never make something that I am not confident to put as my portfolio. But well, everyone have a different standard.
English
2
0
3
287
Jay Yang
Jay Yang@Jayyanginspires·
What are your favorite use cases for open claw?
English
8
0
9
2K
Matthew Turley | Fractional CTO retweetledi
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
Every non-technical founder I talk to is excited to use AI. They make plans, projections, maybe some internal tools and automations, or prototypes for new features. Low risk things where nothing breaks if something goes wrong. Then I ask if they’ve added any new features to their production site with AI. Long pause. No. They are either not technical enough to do it, or technical enough to know that they don’t know enough to trust it. Three founders this month. Same answer. AI makes things easy when nothing is at stake. AI on a live codebase with real revenue is a different job entirely. This is the gap I fill. Agents write the code, I review it before anything touches production. Fine-tune or rework if needed. I pick up when they call on Friday night.
English
1
1
0
67
Paul Mit
Paul Mit@pmitu·
I designed @claudeai t-shirt for myself, what do you think?
Paul Mit tweet media
English
30
1
61
2.3K
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
Not going to get bit by this again. I'm building Trellis, an agent fleet runtime to replace the Claude Code + openclaw setup I've been running. Provider-agnostic, owns its own state, graduates me off vendor-specific CLIs. All running through RelayPlane to optimize token use. Wish me luck.
ClaudeDevs@ClaudeDevs

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK

English
3
1
3
233
Matthew Turley | Fractional CTO retweetledi
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
Anthropic just put a $200/month meter on programmatic Claude. Sam Altman dropped 2 months free Codex in the same hour. I'm not picking sides. I'm routing between them. Cheap models for the boring 70% of agent work, Claude or Codex for what matters.
Matthew Turley | Fractional CTO tweet media
ClaudeDevs@ClaudeDevs

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: - Claude Agent SDK - claude -p - Claude Code GitHub Actions - Third-party apps built on the Agent SDK

English
0
1
1
160
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
@aroido_bigcat Yeah, that's the thing I'm most worried about. Approval rules version like code. Every decision row pins the rule hash + inputs + outcome, so "why did this auto-merge" returns the YAML that actually fired, not whatever's current. Rule edits don't rewrite history.
English
1
0
1
17
bigcat | Aroido
bigcat | Aroido@aroido_bigcat·
@matthewrturley That split makes sense. The decisions themselves are the part I’d want most inspectable. Plans drift, but an invisible approval rule is where the scary bugs hide.
English
1
0
1
13
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
All three, plus the decisions themselves. Workers are stateless on purpose. State lives in the director's SQLite: task plan, approvals, run history, decisions, fleet topology. Authority is per-decision-type YAML. Strict gates at first, then trust-based auto-promotion once a decision type proves itself.
English
1
0
0
31
bigcat | Aroido
bigcat | Aroido@aroido_bigcat·
@matthewrturley Owning state is the boring part until a vendor CLI changes under you. I’d be curious what you keep outside the workers: task plan, approvals, run history, or all three?
English
1
0
3
18
Fed 🐻
Fed 🐻@foliofed·
Realizing prompt engineering isn't that important for controlling AI costs Pipeline architecture is: - cheaper models for low-stakes tasks - smaller inputs - smarter sync cadence the prompt is the last thing to optimize
English
10
0
17
1K
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
@hnshah The shipping-fast part is easy now. The learning from it is where I've been focusing recently. Trying to find the right balance. Somewhere between complete bottleneck and "no idea what's going on".
English
0
0
1
46
Matthew Turley | Fractional CTO
Matthew Turley | Fractional CTO@matthewrturley·
Marc Benioff on All-In (@theallinpod): "an 'intermediary layer' will be needed to decide which inputs go to Anthropic and which go to smaller models." Been shipping this for six months. Auto routes hard reasoning to Opus or GPT-5, the rest to Haiku, Groq, GPT-4o-mini, or a local model. Cuts agent bills 40-60%. $150M back per year npm install @relayplane/proxy @Benioff @chamath @Jason now accepting 1% of year-one savings as seed
The All-In Podcast@theallinpod

“I am going to probably use $300M of Anthropic this year at Salesforce.” - Marc Benioff “ These coding agents are awesome. Anthropic is awesome. Coding, everything's going to be cheaper to make, it's more efficient. I can do things that I just could not do before. I can go faster than ever before. I can implement my software and sell it at the same time. I've never been able to do that before. Today, I have humans, agents, and headless platforms all interoperating, never before. So the opportunity for my own company and the efficiency that I have in my own company, in service and support, in distribution and marketing, across the board, is unprecedented. What I can do for our customers, unprecedented. And, to that point, my gosh, have you seen Anthropic? It is a rocket ship that will not stop.”

English
1
0
1
182

Keşfet

@phuctm97 @stevelauda_ @Jayyanginspires @pmitu @claudeai @aroido_bigcat @elonmusk @BarackObama