CTO & Co-Founder @Faros_AI

San Francisco Bay Area
Joined February 2011
Matthew Tovbin retweeted
I really enjoyed my conversation with @RthvikaS of Bloomberg about the AI bills landing on bank CFOs’ desks. Her piece published today is a sharp read on a turning point most companies haven’t fully reckoned with yet, and I’m glad to have contributed.
1
1
4
182
Matthew Tovbin retweeted
AI has your engineers writing code faster. So why does it feel like everything else slowed down? Join Faros on November 4 for a fireside chat on what our latest research data shows and where to intervene. Register here: faros.ai/events/your-ai-bott…
1
4
121
After seeing this, I bet @sama finally cancelled his order. Sorry @elonmusk
RDSTR - 2017 VS 2028 I‘m not sure what to think of the new design. Rear end looks brutal but the rest? 😵‍💫
56
Matthew Tovbin retweeted
If you are running AI coding on Amazon Bedrock, the usage data is easy to get. What it doesn’t answer is whether the spend changed how much your team ships. We discuss AI outcomes with Jeff Grimes, Head of Global Startup Program at @awscloud: faros.ai/events/optimize-ai-…
1
3
3
102
If you are at the ELC Annual even next week - come talk to us.
One blanket cap for every team and model means a token that ships a deployed feature looks the same as one that goes nowhere. At ELC Annual next week, we'll show how to tell the difference: invest where AI works, reroute where it doesn't. Grab time with us: pages.faros.ai/meet-with-far…
2
3
123
Matthew Tovbin retweeted
Engineering teams that lean harder on AI are seeing more of their work come back. Find out where your work restarts come from and how to cut them with our latest research blog:
1
1
2
243
Matthew Tovbin retweeted
Introducing Token Engineering. Builders are burning huge token budgets, and leadership has no way to see what it delivered. Faros shows what each dollar of token spend actually bought — features shipped, incidents resolved, customer issues closed — at the team and individual level. Is your team's token spend too high, or not high enough? Faros proves which models and context perform best on your own code history, and enforces policy against it — so you can answer that, and make every dollar count. Stop token maxxing. Start outcome maxxing. faros.ai
2
8
15
144
Matthew Tovbin retweeted
Stop by booth #173 to see how engineering leaders are using intelligent tokens to trace AI coding spend to the work it produces, optimize model routes and workflows, and govern it at scale. Book time with our team: pages.faros.ai/meet-with-far…
2
68
Matthew Tovbin retweeted
Model routing needs to account for more than cost and quality. Our recent GLM 5.2 pilot made one capability boundary especially clear. As coding workflows become more complex, routing policies need to become more context-aware too. See why: faros.ai/blog/glm-5-2-claude…
1
1
79
Matthew Tovbin retweeted
That AI coding model that looks cheap per million tokens might actually be the more expensive choice once you count the failed attempts, verification, review, and rework it takes to get usable work out of it. More details here 👉 faros.ai/blog/why-cheaper-ai…
1
2
63
Matthew Tovbin retweeted
This is what happens when you vibe-code a product that handles sensitive user data: I saw this extension announcement going viral and thought it looked cool, so I checked it out. My first concern was whether my emails would be sent to someone I don't know. I was relieved to see you could bring your own Anthropic API key. The website states your API key and email text go straight to Anthropic and "never touch our servers." Then, when you try it, you hit a paywall where you can input your key. Same claim: your key is stored locally and calls go directly to Anthropic. I got curious and decided to check the Chrome extension source code. Damn. Not only does it send your email text to their servers, it also sends your API key! So both claims, that it calls Anthropic directly and that your API key is stored only locally, are completely false. I'm giving the creator the benefit of the doubt and assuming this was negligence, not malice. He used his real name on Stripe and in the Chrome Web Store. But the lesson is bigger than one project: building with AI is fun, but if you're handling people's data, making security claims, and charging money, you need real review before shipping.
I made the anti-Grammarly. Mess up your emails with AI. Sinceerly.com
89
87
27
1,472
238,635
Matthew Tovbin retweeted
2 years of telemetry, 22K devs. AI's impact on engineering: throughput is up, incidents per PR tripled, and 31% more code is shipping with no review at all. That's the Acceleration Whiplash. The AI Engineering Report 2026 is out. faros.ai/research/ai-acceler…
3
3
1
5
311
Matthew Tovbin retweeted
Hey @awscloud, a new API should NOT be a health event. Stop waking people up with spam.
4
17
382
22,768
Matthew Tovbin retweeted
Brave just registered a .agent domain! We support the effort to have the .agent top-level domain managed by a community, instead of being owned by one company. Join the community and pre-register your domain here: agentcommunity.org
128
249
121
2,994
544,171
Matthew Tovbin retweeted
very cool to build on decentralized web3 infrastructure but still have your app go down because of *checks notes* a cloudflare issue in ashburn, virginia
2
1
12
2,231
Matthew Tovbin retweeted
🚀 AI transforming software dev? Get REAL visibility into its impact on productivity & business outcomes. Faros AI is NOW on @Azure Marketplace & MACC eligible! Learn more: faros.ai/blog/faros-ai-partn…[…]al&utm_campaign=2025-pr&utm_content=microsoft-azure @msPartner
1
2
268
RT @Osint613: THESE BARBARIC MONSTERS KIDNAPPED A 10-MONTH-OLD BABY AND A 4-YEAR-OLD CHILD—NOW THEY’RE SENDING THEM BACK IN BODY BAGS. THE…
This quoted post is unavailable.
42
17
Red Cross: Humanitarian bystander of the year ft. @ShutupLyle
315
2,289
490
8,814
861,457
Matthew Tovbin retweeted
Donald Trump: “If the hostages are not released prior to January 20, 2025, there will be ALL HELL TO PAY in the Middle East.”
213
646
48
4,327
77,443