GutFeel Capital

@ShanuMathew93 isn’t this all a year old? Yday was the 28/29 and uncapped would have been $555

@kevindstevens What about not capping prices, to incentivize new build so that prizes go down…

ERCOT catches a lot of heat but PJM is headed for trouble too. This capacity auction hit the cap of $325, but… “Without the cap, prices would have surged to $554.72 per megawatt-day for most of the regional grid, PJM said in a release. In the region covered by electric utility ComEd, which includes parts of Illinois, prices would have cleared at more than $775 per megawatt-day.” This type of spread will discourage new generation investment and that’s only going to cause issues in the long run. I’d be long anything resiliency related or tools / services that keep existing assets going past their intended lives.

@fwikingsson Filip vedervärdig i detta

Minns ni filmen ”Sophie’s Choice”? Nähä, den handlar om en kvinna (Meryl Streep) som typ 1942 tvingas göra ett omöjligt val. Historien upprepar sig på lördag. Inte för Meryl Streep men för mig och många som är som jag. open.spotify.com/episode/2ew3oP…

@NewmoonCap lol, go look at Vlad’s Alstom work over time

Junior staffs looking at Vlad's calls and be like, I can be head too.

>Be Vlad >Head of Barclays Capital Goods Research >Underweight Wärtsilä >Underweight Vestas >Downgrade Siemens Energy >Target prices keep getting steamrolled by the share price >Still shows up acting like the market is the dumb one 1/n

@DeepSailCapital But also why higher margin businesses should also have lower multiples as they mature. 20% margin biz with 80% incrementals and growth runway vs. 60% margin biz with 80% incrementals. Revenue growth turns into profit at 4x vs. 1.5x rate

The higher your margins the less changes in your margins will impact your earnings. That’s why high margins matter to business quality. A high growth high margin company will compound earnings at near the revenue growth rate even with declines in margins.

@BodenPJ Går bra för Dansken!

Dansken på Elekta har chans att bli Sveriges högst uppskattade VD om han levererar en riktig turn-around. Vilka Svenska bolag har haft Danskar som levererat? Mark Jensen på Ambea någon till?

@ShanuMathew93 Moses is the goat

Different time periods so apples-to-oranges but for scale. Compare to Jefferies at: "estimate ~50GW of data center BESS volume through 2035, or ~125GWh." Benchmark via Mizuho: "Benchmark’s ~25-30 GWh US AI/DC BESS view implies ~1-2 hr duration per GW of compute, above our prior ~1.0 GWh/GW buffering framework from Ascend"

@TechFundies Incredibly overpriced relative to alternatives

If you are a $FDS customer, be sure to call your rep. Mine says nearly all clients are trialing the MCP, and 100% of completed trials converted to paying for the connector. 100%. List price for the connector is a 20% uplift to contract value for the first 10 seats and then drops to an incremental 15% for additional seats. Said there has been 0 pushback on price. Single seat pricing still a work in progress as that is a minority of their business. Stock trades at 14x this year's GAAP EPS estimate which is likely too low, and 13x next year which is highly likely to be too low...

@code Not working despite having updated

📣 Claude Opus 4.8 is now rolling out in @code. Give it a try!

@mason_8718 What’s the source?

🇶🇦🇰🇼🇸🇦🇦🇪🇮🇶🇰🇷 Huge news. Qatar and Kuwait are soon set to sign contracts to purchase the M-SAM-II ballistic missile interceptor system reportedly. After the M-SAM-II's success in intercepting an Iranian threats in the UAE and the surge in global demand for air defense systems, LIG D&A has revealed that inquiries are coming other Middle Eastern countries, Asia, Europe, and South America. Currently, South Korea, the UAE, Saudi Arabia, and Iraq have ordered the M-SAM-II in bulk, and in response, missile factory is doubling its production capacity. ●We look forward to news of contracts for systems such as L-SAM, LAMD, and H-SHORAD(Biho-II) in the near future. ●The export version of the M-SAM-II features Hanwha Systems' new MFR-E radar. Also known as SDR (Software Defined Radar), this system utilizes liquid cooling technology in an embedded format to ensure optimal semiconductor performance in prolonged external high-temperature environments, and includes additional features such as hypersonic target tracking. Like other latest models in Hanwha Systems' MFR family, the MFR-E features a main array antenna configured using an advanced TRA (Assembly) method based on GaN technology. (LIG D&A also possesses radar technology at a similar level.) ●South Korea is one of the few countries that possess ballistic missile interception technology at low, medium, and high altitudes and sales it in bulk without political, diplomatic, religious, industrial, or ideological conflicts. ●Most of all, there is nothing more noble than protecting the lives and assets.

@FundamentEdge @AryamanIyer3 @ecommerceshares You are using Claude Chat or Claude Cowork and comparing that to Codex? Hmmm

@AryamanIyer3 @ecommerceshares that may be true in a custom harness (idk). But Codex vs. Claude Chat or Claude Cowork isn't even close right now.

Claude isn’t “more intelligent” necessarily, but it does what none of the other top-end models is able to reliably do: It follows instructions. In finance, that’s really important. If you’ve ever worked with someone who doesn’t follow instructions, you know how bad that is.

@FundamentEdge @ecommerceshares Not true

@ecommerceshares Codex meaningfully better for finance use cases

@TheLAPurchaser Can’t DM you

What does Aiera cost for MCP integrations?

@ecommerceshares Skill issue

I just spent 90 minutes using three frontier models for some financial analysis. Had them check each other. Got endless hallucinations & logic errors. This work would’ve taken me 30 minutes if I had just done it all manually. We’re not even close to AGI, folks. Early innings.

@FundamentEdge Only talking about deep research here. Have been testing with same prompts every month. Got clearly/consistently better results in January now. The biz dev team I talk to there acknowledges but don’t know why happening

@FundamentEdge Using many different models introduces a lot of volatility in quality of the outputs. Degradation in terms of visuals/tables/presentations. Feels like token budgets gotten smaller rather than opposite for other models

I'm building my AI-native mock portfolio on Perplexity Computer for a few reasons: - It's multi-model (for example, GPT 5.4 is much better at auditing excel than Opus 4.6, and Claude models haven't been as good at images). The orchestra seems to work much better. My hypothesis is multi-model wins long-term as it avoids the LLM leapfrogging issue and can route tasks to high intensity (high inference) and low intensity tasks. - For whatever reason, PC is much better and more reliable at "meta Skills" creation and "iterative Skills" improvement than Claude, which is really important for me, given I have 3,000 slides, 306 prompts from my 2025 prompt library & thousands of pages of context on the investment process which i am turning into Skills files (per Karpathy's "knowledge base" approach). I learned the hard way taking 5 prompt engineering courses summer '25, only to never write another prompt by hand...LLMs quickly abstracted prompt engineering and PC has abstracted Skills creation. This is really important for my consulting work where I take a baseline Skills backbone then iterate & customize to the specific investor (almost like setting up a Bloomberg launchpad). It takes just voice record, no engineering. - It isn't buggy. I get almost no "you've reached the limit of tool use" sort of errors. For all it's praise, I find the Claude ecosystem incredibly buggy. And seems to get nerf'd/throttled from time to time. - It doesn't operate on my system, but in a virtual file. I get a little uneasy pointing Claude CoWork to my file system. Not to mention the endless verification loops required if you don't want to unleash Claude Code on your system. - Trusted friends told me it would take me 100+ hours to develop real fluency with coding agents to the degree that I could adequately debug and become a fluent power user. I'm doing this as a guinea pig for my clients who are predominantly 25-2,000+ global investment teams. It's absurd to think an investment team of 100 is going to take the time to use VS Code & learn to debug code. For devs sure, for investors, no. - I do run out of tokens incredibly fast. My credit card is warm. I'm not crowning Perplexity Comptuer the winner, but my hypothesis is that something like this becomes the backbone of the new investment dashboard. To not be perceived as a shill, I have no economic relationship with Perplexity. I am a paying user. The lesson from me as tools has improved is the leverage in Workflow Context (deep articulation of what to do and how to do it) and Research Context (sell-side notes, expert network calls, data filings, transcripts, internal notes etc). The real improvement in AlphaSense outputs over the last 9 months has shown me the leverage in Research Context, and this experimentation in PC has show me the leverage in Workflow Context. The New Investment Dashboard, in my opinion, is powered by Workflow Context and Research Context and has abstracted away the engineering such that the set-up process for the front office fundamental investment professional is <60 minutes.

@mason_8718 Korea and UAE officially LSAM deal coordination?

Aside from the recent situation in Iran, Iran and Israel were firing missiles at each other at will over Iraq just a few months ago. Geographically, Iraq stood at the center of challenges and provocations, so close cooperation with South Korea was only natural in this situation. (To be more precise, the M-SAM-II deal dates back even further.) Just a few days ago, high-ranking Iraqi military officials visited Hyundai Rotem as part of the K2ME tank deal. We expect that Baghdad can be effectively protected with about two L-SAM (Block-I) batteries without incurring a large budget. (Officially, South Korea and the UAE are currently coordinating the L-SAM deal.)

@TheLAPurchaser You dont have Bloomberg? You can connect api to claude

What free plugins/APIs do you use with Claude to get access to historical market data?

@mason_8718 What do you think about the LIG Nex1 and Palantir MOU?

<Hanwha official account> "The K9 is the world's most proven artillery system, already in service with 11 nations, including six NATO countries such as Norway, Poland, Finland, Estonia, Romania, and Turkey." 🇰🇷 🇹🇷 🇵🇱 🇫🇮 🇪🇪 🇮🇳 🇳🇴 🇦🇺 🇪🇬 🇷🇴 🇻🇳 🇪🇸

🇪🇸🤝🇰🇷 Breaking News. Hanwha Aerospace and Indra have officially agreed to partner on Spain's next-generation self-propelled howitzer program. This project is considered a key undertaking for the Spanish military and is estimated to be worth 4.5 billion euros. The project comprises a total of approximately 330 units, including 128 next-generation self-propelled howitzers, 128 ammunition resupply vehicles, 21 recovery vehicles, and 59 command and control vehicles. Spain will lead the platform design and production based on K9, while Hanwha will exclusively supply key components. Also, partnering with Spain makes it easier to target the Latin American market. Meanwhile, Hanwha Aerospace is focusing on the simultaneous development of the next-generation self-propelled howitzer K9A2 and the ultra-long-range system K9A3.