Gerry Burde

[Gerry Burde @GerryBurde]

@kamilbaranek_ Google $20 AI plan is good for images via antigravity harness, chat, and Flow.

[Gerry Burde @GerryBurde]

I have only used Codex Plus and Gemini AI Pro, I think overall value goes to Gemini if you use the storage and other benefits. Both options also have pretty generous free offerings too if you stick with flash/luna. Command Code and OpenCode $10 plans are good competitors too, especially because they have free/preview models available regularly(OpenRouter does too, with no subscription I think you just needed to have funded $10 in the past).

[Gerry Burde @GerryBurde]

@karlclement Integrations is the only reason, Google Jules, OpenAI Dots, etc. It's a whole process getting PRs back to my private git forge though.

[Gerry Burde @GerryBurde]

Ironically, thinking that "reading all the code on the internet" makes an LLM a master software engineer is the actual Dunning-Kruger effect on display.

You are defining engineering as syntax retrieval, boilerplate scaffolding, and memorizing API quirks. Yes, models crush humans at memorization and typing speed.

Real software engineering is:

Knowing which trade-off to make when every textbook design pattern fails.

Navigating undocumented distributed runtime state, race conditions, and business constraints that never lived in a public git repo, or do, but the agent didn't read that file into its context window.

Not letting an agent hallucinate architectural drift because its context window is drowning in noise.

If 100% of your job was writing leetcode one-liners and standard CRUD glue code, sure, you’ve been beaten. But for everyone building non-trivial, mission-critical systems, an agent without an engineer is just an extraordinarily fast way to generate technical debt.

The models might have all the knowledge of the universe, but they can only remember 1m tokens about your specific task. They're excellent for tasks that can be auto-researched and have a clear iteration ladder, but fall apart on larger tasks. Maybe in another 6 months we'll need less human glue.

[Gerry Burde @GerryBurde]

@neogoose_btw @IroncladDev the skill should ship inside the app as a --skill so we can ask our agents to read the docs

[Gerry Burde @GerryBurde]

Ahh yes, accountability... Foreign to SWEs!

For decades, software engineers enjoyed a unique corporate sanctuary: an unspoken agreement that writing code was an artisanal, intangible craft that couldn't possibly be subjected to crude metrics like sales quotas or factory unit economics. Performance reviews were cushioned by the subjective haze of "impact," peer vibes, and debating whitespace in PR reviews while the business waited on deliverables.
​Cognition’s model strips away that sanctuary in one stroke. When high-tier compute is meted out strictly by quality and outcome efficiency, software engineering is finally dragged into the cold light of ROI. If your agent is burning millions of tokens hallucinating in circles, it’s not because the problem is mysteriously complex—it’s an unvarnished diagnostic that you don’t actually know how to define a problem, structure context, or verify an architecture.
​The real existential panic isn’t about token rationing; it’s about outcome accountability. For an industry that built entire careers out of hiding behind the perceived mystique of the development process, discovering that your cognitive clarity is now an observable, measurable balance sheet is the most terrifying performance evaluation imaginable.

[Gerry Burde @GerryBurde]

@ArnoldCartagena @Im_IrushiK Please ask AI why you would opt out of a reset which resets your 7 day timer(It's bad anytime you've used less than 10% per day of your quota).

[Gerry Burde @GerryBurde]

The claim that AI needs all global software revenue plus $1.5 trillion next year (~$2.35 trillion total) to break even on data centers is off by roughly an order of magnitude.

Here is the actual breakdown:

1. The Numbers Don't Add Up
Total Global Software Market: ~$850 billion/year across all enterprise software and SaaS worldwide.

The Claim: $850B + $1.5T = ~$2.35 trillion in annual revenue next year.

Reality: Requiring $2.35T in a single year treats cumulative multi-year infrastructure investments as an immediate, one-year expense rather than depreciated assets.

2. Actual Capex Spend to Date
Hyperscalers (Microsoft, Alphabet, Amazon, Meta) and Tier-2 clouds (Oracle, CoreWeave, xAI) have ramped spend aggressively, but nowhere near that scale:

2023–2026 Cumulative AI Capex: Approximately $550B to $650B total (GPUs, networking, data centers, and power).

Most of this is funded directly out of corporate cash flows (search ads, cloud services, enterprise subscriptions), not debt.

3. Break-Even Math at 9% Interest
Even if we assume the entire buildout was 100% financed through debt at an aggressive 9% annual interest rate:

$500B Capex Base (5-year payback on hardware):
Debt service (principal + interest) = ~$129B / year

$800B Capex Base (aggressive peak buildout, 5-year payback):
Debt service (principal + interest) = ~$206B / year

$800B Capex Base (10-year blended payback for physical facilities/power):
Debt service (principal + interest) = ~$125B / year

The Takeaway

Adding data center power, maintenance, and operating costs brings the annual industry revenue needed to sustainably support this infrastructure to around $350B–$450B per year—not $2.35 trillion.

[Gerry Burde @GerryBurde]

@TechFundies Same reason all the labs are building their own inference chips. keep the margin for themselves. Also look at the market caps.. Meta is $2T, with $70B in profit... Meanwhile OpenAI and Anthropic are both $2T with no profit. They could double their cap if they make SOTA models

[Gerry Burde @GerryBurde]

@GregHBurnham I don't know, I have for all intents unlimited Opus 5.5 / Astra 6 instances in 2026 and it hasn't made me rich yet. One instance would be a bit limiting, but at least a few million by 2020.

[Gerry Burde @GerryBurde]

People are bringing 10 year old GPUs back online for inference... Newer more efficient GPUs aren't replacing older ones.

The "token factory bankruptcy" narrative applies a 19th-century railroad bust or merchant-power model to an asset class operating under entirely different unit economics, capital structures, and physical constraints.

To see why a systemic collapse in token production isn’t materializing, you have to break down three structural realities:

1. Marginal Costs Are Not Stranded by Power or Facilities
In traditional commodity collapses (like natural gas or merchant power plants), operations halt because marginal cash operating expenses (fuel, facility rent, turbine maintenance) eclipse the market clearing price.

Inference doesn't work that way. Datacenter rack space, cooling, and wholesale electricity account for a single-digit percentage of total token delivery cost. When the marginal cost to keep a chip powered is practically negligible compared to its capital cost:

Operators don't shut chips off when prices drop—they bid them into lower-tier inference tasks.

Even decade-old hardware (such as legacy Pascals or early Tensor Core architectures, V100's etc) is actively racked and profitable today for low-throughput, batch, or non-latency-critical serving.

There is no operational cash drain forcing operators into Chapter 11 shutdowns simply because token spot prices dip.

2. Balance Sheets Are Captive, Not Levered Merchant Foundries
A traditional capex bubble pops when independent producers finance hardware with high-yield debt, banking on fixed forward contracts that get broken.

The vast majority of hyperscale token production is funded directly out of the balance sheets of cash-flow machines—enterprise software, search monopolies, and advertising platforms running 30–40% operating margins:

These entities do not face bank-driven liquidation cycles.

If hardware becomes economically obsolete before its scheduled accounting depreciation, it gets written down on the balance sheet as an impairment, not liquidated in a bankruptcy auction.

Captive internal demand (search enhancement, coding assistants, internal tooling, ad targeting models) absorbs capacity even if external API prices crater.

3. Autoregressive Inference Is a Memory Wall, Not a Compute Glut
The core premise of the commodity-bro thesis is that flooded silicon capacity drives token value to zero. But inference is not compute-bound; it is strictly memory-bandwidth bound.

The Generation Chokepoint: During autoregressive generation (token-by-token decoding), the model must stream entire weight matrices through memory just to produce a single token (arithmetic intensity near 1 FLOP/byte).

The HBM Moat: A glut of raw tensor FLOPS does not crash token supply if memory bandwidth remains scarce. High Bandwidth Memory (HBM) packaging and interconnect fabric are the actual bottleneck.

The Floor on VRAM: Because inference throughput is dictated by gigabytes-per-second of bandwidth rather than raw compute cycles, older high-bandwidth silicon continues to command a bid. Every unit of memory bandwidth that can hold model weights and the KV cache generates revenue.

A systemic bust would require an order-of-magnitude architectural disruption in memory bandwidth efficiency—such as radical sparsification, 1-bit quantization without perplexity penalties, or non-autoregressive architectures that break the memory-streaming constraint entirely. Until that physical bottleneck disappears, memory scarcity establishes a structural pricing floor that prevents an unconstrained supply collapse.

[Gerry Burde @GerryBurde]

@_FORAB There will be nearly unlimited demand for DDR5 if prices for it were back at 2024 levels. We don't need HBM for inference, it's just a nice to have. The incumbents should be scaling all of their manufacturing to keep market share, but CAPEX barely seems to have accelerated all

[Gerry Burde @GerryBurde]

It's not because it's hot, it's because 62,000MW of capacity is offline for maintenance(which is scheduled for this cooler period of time..). This notice isn't an actual emergency, it's just the method used to allow heavy industrial users to switch to their on site power without constraints

[Gerry Burde @GerryBurde]

If Maia is in fact 1.7x more efficient than GPU, and they produce them for cheaper than the NVIDIA tax, I'm sure it'll reduce demand for private inference by the fortune 500 and other companies/CoPilot users that trust Microsoft or have other reasons to pay a premium for hosted inference. Eventually the default should become private inference, but the TPU's or other solutions need to actually be available and not in a walled garden of hyperscalers

[Gerry Burde @GerryBurde]

@LeachartsMarte @HannoKlausmeier What makes it expensive? Batteries? Have you considered they will be 90% cheaper in 5 years? Electricity? Have you considered you can make your own for 5c/kwh in the back 40? Is it the reduced maintenance?

[Gerry Burde @GerryBurde]

@BridgeObsidian @satyanadella Hyperscalers wouldn't, but someone should be building these for the several trillion dollar enterprise and consumer market.

[Gerry Burde @GerryBurde]

It's most likely the opposite. More people will go to restaurants because the quality will(can) be higher due to shorter inventory cycles, it can be more varied due to larger inventory of spices and ingredients, higher quality cooking implements, and cheaper procurement of ingredients. People eat at home because it's cheaper and they get exactly what they want. Neither of those need be exclusive to home meals.

[Gerry Burde @GerryBurde]

@wertform_mcm I bet it's 2x better

[Gerry Burde @GerryBurde]

@allTheYud Maybe even Late November without something groundbreaking

[Gerry Burde @GerryBurde]

But still $110 Billion of Office licenses are sold every year.

Microsoft 365 pulls in $110B+ annually with over 400M paid commercial seats. As a standalone company, that’s a ~$1.2T+ valuation.

Good luck getting corporate legal to redline contracts in Git, or enterprise auditors to debug a Python script instead of an Excel model. Office isn't just doc tools—it's the identity, compliance, and communication operating system of the Fortune 500. Can other solutions replace that stack? Probably.