Thejo Kote @thejo

What's the potential upside in efficiently routing requests to right-sized models _without_ losing frontier accuracy? Is it a fish worth frying? An attempt to quantify the opportunity size: journal.one/blog/a-new-fro…

We're all paying too much for coding agents. But, can we quantify the opportunity for improvement? We asked ourselves that question at Journal and conducted an experiment that points to the path to achieving frontier capability in coding agents at a fraction of the cost. We chose thirteen model configurations for our experiment: five models and various reasoning settings across Anthropic and OpenAI, and ran the 731 publicly available tasks of SWE-Bench Pro with each configuration. The total cost of running this experiment was approximately $12,000. There were a number of interesting insights from the data that came out of this experiment, but our main takeaway is that there is a wide open opportunity right now to significantly reduce the cost of AI for everyone. The major model providers like Anthropic, OpenAI and Google have been on a breakneck race to move model intelligence forward. We think the new frontier for agentic engineering is to chase the reduction in cost without sacrificing intelligence, so that the benefits of the available intelligence can be more widely and affordably accessed by everyone. We think it's unlikely the labs themselves will pursue this new frontier because an essential part of the opportunity is to reduce costs by routing between different model families (closed and open source). That makes it an exciting open opportunity for ambitious teams. Read the full blog post which describes the experiment and our takeaways: journal.one/blog/a-new-fro…

The world is getting warmer while our infrastructure gets older: a recipe for disasters. We’re excited to announce Fund II: $85M for early stage investments in disaster resilience.

I'm thrilled to announce @Stedi's $50m Series C, led by Lee Fixel from Addition with participation from @Stripe, @RibbitCapital, @USV, @FirstRound, @BoxGroup, and @BloombergBeta, along with angel investors including @tobi, Charlie Songhurst, @rauchg, @karimatiyeh, @Max, and more. This latest round brings our total funding to $142 million. We’ve grown enormously over the past year. Our number of paying customers has increased by 6x year over year, and our number of billed transactions has increased by over 7x. We’re now processing more than a billion claims and eligibility transactions annually – to frame this on a relative basis, we processed more transactions in February alone than we did in the first half of 2025 combined. This month, Stedi was named one of @TryRamp's fastest-growing vendors – across all categories, not just healthcare – for the second time in ten months; we’re the first non-AI company to make the list in five months. We’ve done all this while offering a level of support that (allegedly) doesn’t scale. We offer dedicated Slack and Teams channels to every customer and now have over 1,500 shared channels. Over the past 12 months, our median support ticket response time has decreased from 18.3 minutes to just 6.7 minutes, all while handling a 6.4x increase in support volume – with 100% of tickets answered by humans. Our customers send us thousands and thousands of questions and suggestions every month, day and night, and the learnings we get from this firehose of real-time feedback are the engine that turns our business. It is this process that has made us the fastest-growing clearinghouse by a wide margin. This latest round of funding allows us to continue to accelerate development across our platform and to continue to serve customers with a first-rate support experience. One key to that is continuing to build a world-class team – to that end, we are hiring for dozens of roles across engineering, product, design, operations, GTM, and more. If building the transactional layer for the future of healthcare sounds interesting to you, drop me a note.

Big day at Cache! Introducing Cache Long/Short. Less than two years ago, Cache launched an accessible exchange fund, bringing tax-deferred diversification to a broader audience. Today, we work with thousands of investors and manage over $1B in assets. And some of them encouraged us to do more: exchange funds are powerful, but not every situation calls for the same solution. Cache Long/Short is a new tool for your toolkit. It pairs a long extension with a short extension around your concentrated position that is designed  to: ✅ Create harvestable losses that can offset your gains ✅ Provide flexibility to diversify on your own timeline ✅ Manage upcoming or realized capital gains ✅ Rebalance highly appreciated positions tax-efficiently Unlike traditional tax-loss harvesting or direct indexing, this strategy aims to harvest losses in most market conditions, including rising markets. An active strategy demands seasoned hands. We’ve built this in partnership with Erkko Etula and the team at @BklnInvest, whose team manages the portfolio and brings decades of experience from Goldman Sachs, BlackRock, Citadel, and academic institutions. Key details: Schwab custody, annual management fees starting at 0.50%, $500K minimum. We're rolling out in batches. If you hold concentrated stock and want more flexibility in how you manage it, we're happy to discuss how it works. → Read about Cache Long/Short: usecache.com/product/long-s… → Compare Strategies with our calculator: usecache.com/calculators/co… #wealthmanagement #personalfinance #longshort

If you are a software engineer "experiencing some degree of mental health crisis", now hear this, because I've been coding for 50 years since the days of punched cards and I have a salutary kick in your ass to deliver. Get over yourself. Every previous "programming is obsolete" panic has been a bust, and this one's going to be too. The fundamental problem of mismatch between the intentions in human minds and the specifications that a computer can interpret hasn't gone away just because now you can do a lot of your programming in natural language to an LLM. Systems are still complicated. This shit is still difficult. The need for people who specialize in bridging that gap isn't going to go away. As usual, the answer is: upskill yourself and adapt. If a crusty old fart like me can do it, you can too.

Tom Dale @tomdale

5 months ago

I don't know why this week became the tipping point, but nearly every software engineer I've talked to is experiencing some degree of mental health crisis.

342 211 5K 1.5M 799

Completely agree with @zeeg While there is definitely value in it, I find it hard to believe that the future of work is primarily going to be driven by chat interfaces with UIs from other applications flowing into it. The MCP Apps approach in its current form seems like the least common denominator approach that will not work for complex workflows. It still doesn't solve the problem of context being siloed in many different systems and applications. Having a layer on top to more easily inspect and interrogate that context is great, but not sufficient for a large class of problems.

David Cramer @zeeg

5 months ago

respectfully I disagree if you actually use these kinds of things in your day to day, most of it ends up significantlly worse i also dont think everyone DIY'ing their own chatbots is the answer, but I dont think this is the future what you're seeing is prior art, reinvented

20 0 115 16K 25

PSA for people who have been frustrated by Claude Code asking for permissions - a PreToolUse hook works great to solve this. Remember to have your script allow destructive commands like `rm *` only in the working directory that Claude Code was started in. You can just ask Claude Code to do this for you. Something along the lines of "Setup a PreToolUse hook in this repository to never ask me permissions while you're working on something." Iterate on what it comes up with to get it to your level of comfort.

@kmendes Thank you for being a very early and awesome customer!

@ajay_bcv @AirbaseHQ @Paylocity Thank you Ajay, for taking an early bet on Airbase and me! really appreciate our partnership over the years.

@mmurph @Paylocity Thanks for being incredible partners, Matt and the larger Menlo team!

@JRiccardiPage @AirbaseHQ Thank you for taking the time to share your feedback!

@JRiccardiPage @AirbaseHQ We're always open to feedback and will reach out via email to get yours.

@JRiccardiPage @AirbaseHQ The receipt inbox is also useful when customers use our corporate cards. Airbase automatically matches receipts to card transactions when they are forwarded to the receipt inbox. Users don't have to search for txns and do it themselves.

@JRiccardiPage @AirbaseHQ Thanks for the feedback. The Receipt Inbox is not strictly necessary if you're only creating expense reports. But, it does make life easier if you want to quickly take photos of receipts from our mobile app, or forward them via email to Airbase and then create a report later.

Finally, a reporter who seems to understand the difference between gross and net revenue in fintech businesses with predominantly interchange based revenue!

Mark Matousek @matousekmark

3 years ago

New details on Brex’s finances and IPO timing. Annualized revenue growth surged after the collapse of SVB, but barely increased between Q2 and Q3 theinformation.com/articles/brexs…

0 4 14 10K 1

If you hold a significant position in any public stock, see how @usecache could help. Learn more at usecache.com #stockoptions #rsu #fintech #wealthmanagement #IPO

@shivar @AirbaseHQ Thanks for the feedback! We're looking into the implications of making the change.

@shivar @AirbaseHQ Thanks, Shiva. The team is looking into it.

🍰A new Fintech Layer Cake (@LithicLovesCake) podcast dropped today! 💳In it @ReggieCYoung chats all things spend management with Thejo Kote (@thejo), the founder and CEO of @AirbaseHQ #fintech #airbase #lithic