Research and compare products on eBay with Kilby.

We rolled out a shopping assistant purpose-built for eBay that enables Kilby users to surface, research and compare eBay listings from within Kilby.

The next feature Kilby will roll out for the eBay shopping assistant will be to enable Kilby users to complete the transaction from within Kilby’s chat. At present, transactions must be completed from within eBay’s checkout.

Until then, happy eBay shopping on Kilby.

The table below is a bit of an eye test, but what it shows is that EV/FCF multiples are not cheap.

The risk to the chip, cloud and frontier lab companies is open source models, as the economics largely go away under open source.

The open source model companies are primarily Chinese companies. American chip companies do not enjoy the same market share in China - a negative for NVDA, AMD, AVGO and MU.

Similarly, companies can host their own open source models without having to rely so heavily on AWS, Azure and Google Cloud to serve up those models.

When we talk about Anthropic and OpenAI versus open source models, it is important to remember that Anthropic and OpenAI have built harnesses, similar to what we have done with Kilby. Claude and GPT are base layer models plus tools. The open source models lack the tools layer, so depending upon what tools you may need, you would have to build the tools yourself should you choose the open source model route.

Bloomberg, Factset and Refinitiv ought to re-write their codebase to incorporate open source models (essentially build APIs off of every product so that models may ingest the data).

For example, if Kilby had a database of 10,000 EPS call transcripts, Kilby would train an open source model on those transcripts and replace Anthropic for Kilby’s /debrief function, which today costs the user $0.25 per query, and could take that cost down to effectively zero.

In a perfect open source world, Kilby could take AI query costs down to zero, and would only charge users a small fee for file storage, which would effectively make Kilby a low cost provider of hyper-personalized, frontier-level intelligence (today for Investment Management and general queries, in the near future Kilby will house other Industry-specific functionality).

Bloomberg, Factset and Refinitiv will not be able to follow Kilby on our low-cost journey because it would cannibalize their business and destroy each company’s revenue line.

So hang tight, Kilby has a bright future with open source models as do other AI-native companies. Incumbents do not.

By the way, here is a copy of Kilby’s debrief output for META’s most recent EPS call complete with Bull Signals, Bear Signals, Deflections and Follow-Up Questions:

• Original META Transcript: HERE

• Kilby’s META Debrief: HERE

Kilby Web: HERE

App Store: iOS app coming soon.

Whomever is loading up on ACN these past few days is making an enormous mistake.

Kilby is coming to the Google Play Store. Try Kilby today: HERE

If you know GUNTHER WERKS, Singer or Tuthill, you know Kilby. Kilby does for Anthropic’s Claude family of models what those restomod firms do for Porsche. Kilby optimizes Sonnet 5 and Opus 4.8 for the Investment Management industry through the application of our proprietary tooling, advanced memory techniques, and industry specific workflows. Kilby is truly unique within the Investment Management industry.

We built Kilby such that it is ultra personal and ultra private. Kilby learns the context of your work based upon your ticker watchlist, your queries, and advanced memory techniques. Kilby stores a record of the query, but not the content of the user’s prompt. We track model usage at the prompt level so that users have maximum transparency into usage and usage-related fees with the security of not having their prompts bleed into the public domain.

As Kilby gets to know you, it will surface ideas, reminders, and opportunities, whether you are a Portfolio Manager, Equity Trader, Research Analyst, I-Banker, or General Operating Executive. Different roles require different types of support - Kilby auto-customizes its support based upon the context of your role, which you help shape during the on-boarding process.

You may also shape how Kilby engages with you from how it responds (in-depth, concise, medium length), the tone of its responses, the sound of its voice and more.

Kilby will be the smartest colleague you ever had. No job is too big or too small.

• Those NVIDIA (NVDA), filings, transcripts and personal notes you have been meaning to review? Upload the files to Kilby and run the “skuttlebutt” command, within minutes Kilby will synthesize all of the files into a cohesive debrief that you may share with your team.

• That transcript you have been meaning to skim through? Upload it to Kilby and run the “debrief” command. Within 120 seconds Kilby will summarize and analyze the transcript, highlight bullish and bearish sentiment, call out areas of the call where management was evasive and offer suggestions of questions to follow-up with management on.

• Have Kilby set price alerts on your watchlist.

• You can ask Kilby anything - it does not have to be investment-related.

  ---

Here are some common Capital Markets use cases:

Market Data & Company Research

• Stock quotes, fundamentals, financial statements

• Company analysis and competitive positioning

• Sector and industry comparisons

• SEC filings (10-K, 10-Q, 8-K, proxy) via EDGAR

• CEO and management background

Web Research

• Real-time web search for news, events, current info

• Fetch investor relations materials, earnings presentations

• YouTube search for conference talks

Analysis & Computation

• Python execution for financial modeling and data transformation

• Spreadsheet creation and PDF table extraction

• Chart generation

Scheduling & Alerts

• Stock price alerts with customizable thresholds

• Scheduled briefs and recurring jobs

Communication

• Slack integration — read channels, post messages

Try Kilby today: https://agentkilby.com/

Kimi K3 is the latest frontier model from Moonshot AI. While the model is free to download, it is not free to run given its size.

For example, running K3 on OpenRouter costs $3/$15 for 1M input/output tokens - less expensive than than Anthropic’s Claude models, but not exactly free.

Moonshot AI prices its K3 model at zero cost, but charges for infrastructure. My view is that $3 and $15 is more about Moonshot AI positioning Kimi K3 at the frontier rather than Moonshot working to recoup OpEx.

I won’t spend time testing K3 as it does not make sense for me to invest in building a tools layer such that K3 would work with Kilby if in fact I can’t materially reduce Kilby’s COGS.

Larger firms will use Kimi K3 as they have infrastructure that I do not, therefore they can run K3 effectively at zero cost. For those larger organizations, the costs stem from calibrating K3’s model weights and having to allocate compute and power to K3.

Yes, you can build real software applications with AI tools. I did it with Kilby, TEK2day Finance, CEORater and more.

Yes, you may also replicate a CRM system as articulated below.

I fear the equity market has been permanently distorted by the Fiscal and Monetary support that has been in place since March of 2020.

“GDP is healthy” the bulls say. Really? Nominal GDP is up as prices have skyrocketed since 2020. “Real” GDP is a fake measure because the Government intentionally understates CPI through two techniques - substitution and Hedonics, particularly the latter. Social security and other welfare payouts are tied to CPI growth, therefore it is in the Government’s interest to understate CPI, thereby tamping down welfare payment outflows.

Inflation used to mean growth in the money supply until the Federal Reserve changed that definition. M1 is growing at approximately 6% year-over-year, much too fast as credit standards are far too lax given the absence of a reserve requirement. The Fed retorts that a zero percent reserve requirement is fine because the market will impose discipline on lax bank credit standards. The problem is that the Fed sets an arbitrary Fed Funds Rate. If the Fed allowed the market to set an overnight rate, and if the Fed did not intervene in credit and equity markets, then under those conditions a zero percent reserve requirement would work. However, that is not the case, the capital markets are not free markets, they are heavily influenced by interventionist monetary and fiscal policy.

In the absence of an honest CPI measure, subtract year-over-year M1 growth from Nominal GDP to get an idea of Real GDP, which is to say that Real GDP is negative and has been negative for quite some time.

Today investors are cheering a favorable PPI read. However, has anyone noticed where the price of oil is? Oil prices have finally started to reflect the reality on the ground and has more upside from here. It is fair to expect that the “Energy” category will have a much higher CPI read in July and that total CPI will move higher for the month.

Equities ought to be trading down, but I’ve never seen a more complacent market with a more complacent investor cohort. What the heck are investors doing all day? I am talking about those of you that have a big smile on your face and think everything is hunky dory with the economy, valuations, and the world.

The rest are like me - more than skeptical - and wondering when the “Everything bubble” will burst.

Capital is walking away from datacenter deals. SpaceX and META are renting GPUs instead of using them. This tells you that AI end-user demand is starting to fall off (although I will use more heavily over time as I bring on people).

By the way, Trump subsidizing gasoline (Freedom Fuel), is another sign the punch bowl will be taken away soon.

Last word, did anyone notice yesterday that Bank of America (BAC) has $85 billion in unrealized losses on Debt securities? We started writing about that loss position in 2022 and the crap leadership team at BofA STILL has not dealt with that situation and the huge opportunity cost associated with that loss position.

Sentiment Analysis for JPM.

• This is Kilby’s “debrief” function, which analyzes EPS call transcripts and returns an analysis in PDF form. Free sample for JPM: HERE. This product is live.

• Check out our earnings call sentiment dashboard (this product is in BETA), which provides real-time sentiment scores for each section of an EPS call, in this case for JPM: HERE.

When IBM whiffed today their pre-announce commentary spoke to a larger issue, which is to say that AI is consuming technology budgets.

Companies are looking at their technology vendor spend and if the organization can’t calculate the ROI generated from using a particular vendor, spend with that vendor is a candidate for being reallocated in part to AI.

I think AI spend will change over the next few years. You won’t see engineers token maxxing, but business users will be expected to use the various tools such as Kilby, Claude and GPT. For anyone who has used the AI tools, you know the ROI is significant. Therefore I expect companies to continue to push usage, but more so at the business user level.

I would not want to lead a large legacy organization today and have to try to convince a legacy employee base to use AI tools. Much better to be an AI-centric startup where employees recognize AI as a force multiplier.

For example, I would rather have two full time marketers who are comfortable using AI tools than a team of 10 marketers who are AI illiterate, and are therefore dependent upon over-priced marketing agencies.

Kilby update: we are building backend technology so that users will have even more detail around their token usage and related costs. We want to be as transparent as possible. The spending insights will be even more powerful as we onboard open source models toward the end of the year.

I expect for Kilby to be in the Google Play store and in the App Store soon. Google Play in August, App Store in September. In the meantime, our desktop and mobile web versions are world class: HERE.

Kilby is optimized for the Investment Management industry, but you may ask it anything.

Here are some of TEK2day’s tools that can help you prepare for earnings season.

For earnings estimates, financials, market data, news, comp tables and more, try TEK2day Finance.

Try Kilby - our AI Assistant for Investment Management and broadly-defined Financial Services (and more). Kilby can analyze and summarize transcripts, financial filings, analyze and build Excel spreadsheets - it can do all of this simultaneously. Kilby is the smartest, fastest, most responsive colleague you’ve ever had.

See an example of Kilby’s earnings call debrief for NVDA HERE.

Visit TEK2day’s Earnings Calendar.

See the full TEK2day Holdings portfolio HERE.

Bloomberg is finally catching on, reporting that the U.S. may subsidize frontier models. No kidding. I think we first mentioned that was coming over a year ago. Others were earlier than us to mention that fact. State subsidy was part of the regulatory capture strategy by Anthropic and OpenAI.

When Claude and GPT are fully entrenched in Govt. operations (they already are), and all user data is cached (Fable/Mythos already caches everything from every user for 30 days, I’m not sure about GPT 5.6), Enterprise customers will balk. They won’t want Anthropic nor OpenAI caching their proprietary data and IP, much less the NSA, DoD and IRS doing the same.

Enterprise customers will walk en masse once they figure out that the model companies are stealing their IP (many already have figured this out), and will move to open source models if they haven’t already.

Anthropic and OpenAI will have 90% of their Revenue coming from Government in 3-5 years if they don’t stop caching and stealing.

The U.S. can only play the SPR draw down game so much longer before that buffer is gone and the price of oil and gas at the pump spike. Iran knows this. The U.S. has zero leverage, so I am not sure why we pretend we do. “Pretend and extend” works in commercial real estate, not in the oil game when demand persists and supply is low.

The risk we run with this pretend game is that Iran will use its obvious leverage, the SPR will be drawn down to a level where we may not draw any further, then Trump would likely want to subsidize the price of oil and gasoline at the pump because he is a big spending politician at heart.

Spending money the U.S. does not have to subsidize the price of oil will lead to more price inflation / a devalued Dollar, and gasoline shortages at the pump. That of course would be Socialist policy not unlike New York City Mayor Mamdani.

Socialism of course does not work. Washington D.C. should read a history book. Better yet, look at how our Socialist policies - Social Security, Medicare, Medicaid, income assistance, a bloated Dept of Education, etc. have destroyed the value of the Dollar.

But then again, there is so much graft involved in these fiscal outflows that benefit members of both parties (my guess is that all members of Congress benefit, especially the Senate, to say nothing of the unelected officials in DoD, State Dept., CIA, IRS, that grow their portfolio every year) that we can’t expect Washington D.C. to police itself. Especially when the U.S. population is doom scrolling rather than paying attention to our fiscal situation.

Time for Americans to throw party politics out the door and to put the Federal Government on a radical starvation diet. It is the only way to make the Dollar and the U.S. standard of living healthy again.

I’ve seen a number of macro analysts parrot concern around the private credit market. It’s true, many of these private credit shops own high yield/junk credit and those underlying businesses are in trouble. The media has painted private creditors as owning software debt and not much else, but that’s not true. Private creditors own strip malls (yes, America’s pizza shops, tanning salons, auto repair shops and local restaurant chains have private debt in those deals), minor league sports teams (even professional teams, we just saw the Celtics trade one of their 2 best players as the PE owner did not want to sign a max extension as he does not own the arena his team plays in). The 10YR has been hanging around 4.5% since April and now with the oil market having cleared that tranche of oil that was floating in the Strait for months, we are seeing the price of oil tick back up, which means upward pressure on yields, which means the screws are tightening on private creditors and the businesses they invest in.

Don’t worry, AI CapEx will support the private credit market - says the media. Will it? Open source models have been taking share for a year and that is only accelerating since the Trump Admin yanked Fable 5. It does not matter that Fable 5 is back, the damage is done.

China is flooding the market with frontier quality open source models. Companies are using those models, they are hosted locally on a private cloud. Those companies control their own data and the model’s weights. Nothing goes back to China. The China fear mongering is a marketing tactic by Anthropic and OpenAI.

Therefore, as China floods the market with open source models, those models will slowly hollow out the economics of the American AI ecosystem. If you don’t need bleeding edge models for your use case, why pay 10x for Opus 4.8 or 20x for Fable 5 when you can use GLM 5.2? There will obviously be a cascading effect to chip prices, including memory, and of course this will have a negative impact on data center builds, at least until the economics sort themselves out. Prices need to come down. For everything.

Money Supply (M1) continues to grow coming off of the 360% growth (Jan. 2020-Apr.2022) during the COVID spending debacle. There are a number of reasons for the money supply growth, none of which are good, all of which I’ve covered in great detail over the past 5 years.

What the graph below tells us is that as the money supply grows, a smaller percentage of the population is working - labor force participation rate is shrinking. So, as goods and services become more expensive (as the Dollar loses value due to the growth in M1), fewer people will be able to afford “stuff” - meaning both essential and discretionary goods and services.

There is a populist story to be told: how the Government spends trillions to satisfy institutional donors, including military contractors. How the Fed subsidizes banks and industry with low rates, QE, various one-off lending programs (ORR, BTFP), and a zero percent reserve requirement (the big daddy bailout which nobody talks about) - industry which has done little to improve GDP (PE firms of course being the big borrowers). Since 2011 PE firms’ key accomplishment is having traded software companies while cutting heads - which does not equal productivity gains - and whilst paying themselves huge dividends. Is that a business model? Seems more like left pocket right pocket reallocation of taxpayer wealth (taxpayer deposits underwrite bank loans to PE firms) to private equity firms.

The answer? Take away the punch bowl:

• Jack up the bank reserve requirement from zero to 10 percent,

• allow Fed Funds to float,

• outlaw QE,

• audit the Fed,

• define a waterfall wind down for Social Security, Medicare, Medicaid and eliminate the related taxes.

• Kill or radically reduce spending on the Dept of Education.

• Break up Lockheed, RTX, Boeing, Northrup, Google, Amazon, Microsoft, Meta.

• Kill the income tax and institute a 10% national sales tax on finished goods and services. This will yield approximately $3 trillion in tax revenue. You could run my skinny version of the Federal Government on $3 trillion and have money to return to the people.

What politician running for President in 2028 is willing to do the above? Instead, the opposite will happen. More handouts, more deficits, more printing, thus goods and services will become increasingly expensive and even fewer people will be working.

This is shitty way to run a country.

ServiceNow (NOW) and Palantir (PLTR) are SIs with software wrappers. Nothing more. If Accenture (ACN), IBM (IBM) and the rest of the legacy SI gang wish to have PLTR or NOW-like multiples, they need to build that infrastructure software layer that allows PLTR and NOW to position themselves as Software companies and charge ridiculous prices.

The reason the legacy SIs are where they are is that none of those companies are led by smart, quality CEOs. I hate to be punitive, but that is the truth. Most are led by weak, unimaginative, non-technical CEOs. I would bet that even Arvind Krishna, IBM’s CEO who does have a technical background, has less than 1,000 hours logged with LLMs. I won’t touch on ACN’s CEO because it was apparent she was a fraud from Day 1. Talk about a WEAK resume for a Technology service provider CEO.

Prompt to Kilby: How would you describe Steve Jobs’ leadership style?

Kilby’s response: Jobs’ leadership style was fundamentally about refusing to compromise on standards — in people, in product, and in taste. Three things defined it:

Obsessive selectivity on talent. He didn’t believe in “good enough” hires. He talked about the gap between the best and average as 2:1 in most jobs but 25-to-50:1 in creative and technical work — which meant hiring wasn’t administrative, it was strategic. The implication is that it is better to have a small team of exceptional people than a large team of adequate ones — he ran Apple’s product line with a senior team small enough to fit in one room. The dark side of this: A players will not tolerate B players — hire one B player into an A team and the team doesn’t get slightly worse, it starts losing its A players who refuse to work alongside people who don’t meet the bar. That’s a real organizational risk most leaders underweight — one bad hire in a high-caliber team can trigger an exodus, not just a performance dip.

Product as the intersection of engineering and humanities.
Jobs believed technology alone is not enough — that it’s technology married with liberal arts that produces work that resonates emotionally.

This wasn’t branding, it was org design:
Apple deliberately employed musicians, artists, historians, and writers alongside engineers, on the premise that engineers alone will solve the problem as defined by the technology rather than as experienced by the person. Practically, this showed up in details like the original iPhone’s slide-to-unlock, a metaphor drawn from industrial design, not software architecture.

Control and simplicity as discipline. He said no relentlessly — to product lines, features, partnerships — because he believed focus was a forcing function for quality, not a constraint on it. When he returned to Apple in 1997, he cut the product line from dozens of SKUs to four. The tradeoff worth being honest about: this style produced some of the best consumer products of the modern era, but it was also famously brutal on people — public dressings-down, binary “genius or bozo” judgments, low tolerance for anything he considered mediocre. It’s a leadership style that works spectacularly when you can actually attract and retain A players who can absorb that intensity. It fails badly in organizations that can’t recruit at that level or don’t have a founder’s authority to enforce it — the harshness without the talent magnet just produces churn.

Prompt to Kilby: How does Job’s leadership style compare and contrast with Tim Cook’s leadership style?

Kilby’s response: Where Jobs was a product visionary who ran on instinct and confrontation, Cook is an operator who runs on process and discipline — and the contrast maps almost perfectly onto Horowitz’s peacetime/wartime framework, even though neither man used that language.

Decision-making style. Jobs made calls by taste and conviction, often reversing direction on a whim if his instinct changed — the famous binary “genius or shit” verdicts, delivered in the room, immediately. Cook is data-first and consensus-oriented — he’s known for going quiet in meetings, asking a barrage of clarifying questions, and only committing once the numbers are airtight. Jobs decided fast and sometimes wrong. Cook decides slower and rarely wrong, but also rarely swings for anything as risky as the original iPhone bet.

What each optimizes. Jobs optimized for the product being right — he’d blow up a shipping schedule over a rounded corner. Cook optimized for the machine being efficient — his background is operations and supply chain (he came from Compaq and IBM before Apple), and his signature achievement pre-CEO was collapsing Apple’s inventory and manufacturing footprint into what’s arguably the tightest supply chain in consumer electronics. Jobs would ask “is this insanely great?” Cook asks “can we build 80 million of these on time, at margin, without a defect spike?” Different questions, both essential — but they’re sequential, not substitutable. You need Jobs’ question answered before Cook’s matters.

Temperament. Jobs led through intensity — public dressing-downs, mercurial mood, an environment that was inspiring for A-players and brutal for everyone else. Cook is famously even-keeled, described by people who’ve worked with him as almost unreadable, calm to the point of unsettling. He doesn’t perform urgency; he manufactures it structurally, through relentless operational cadence rather than emotional pressure.

Risk posture. Jobs was a wartime founder who bet the company repeatedly — on the Mac, on his own return in ‘97, on the iPhone cannibalizing the iPod. Cook has run Apple almost entirely in peacetime: dominant market position, expanding the core (Services, wearables, buybacks) rather than making existential bets. That’s not a knock — it’s the correct mode for the situation he inherited. Using Horowitz’s frame, Jobs was structurally a wartime CEO forced to also do peacetime; Cook is a peacetime CEO who has never really been tested by an existential threat to Apple’s core business. The open question — and the one an investor should actually care about — is whether Cook has a wartime gear if one is ever needed. He hasn’t had to prove it.

What each protected. Jobs protected the product’s soul — simplicity, taste, saying no to nearly everything. Cook has protected the margin structure and capital discipline — Apple’s gross margins, cash generation, and shareholder returns have been remarkably stable and arguably improved under his tenure, even as the “insanely great” cadence of new categories has slowed (nothing since the Watch and AirPods has been transformative on the scale of iPhone).

The honest verdict for an investor: Jobs built the asset. Cook has been an exceptional steward and monetizer of that asset — better at capital allocation and operational execution than Jobs ever was or claimed to be. But Cook has not yet had to demonstrate he can create the next Jobs-scale bet, and Apple’s stock narrative today is more about services margin expansion and buybacks than product breakthrough. That’s a fine business to own. It’s a different bet than owning Apple in 2001.

• $10 input, $50 output per MTok.

• Anthropic model pricing: HERE

• Anthropic really needs to move up the stack: HERE

META deciding to sell compute capacity is akin to xAI renting GPUs to Anthropic and Google (those announcements occurred over the past few weeks in case you missed them).

META is an LLM upstart, it can’t hope to be successful building a leading frontier model and a cloud business to serve up AI infra. It has to pick one.

By moving forward with the AI infra business, META is essentially quitting the frontier model business. META should have stayed focus on open source LLMs.