chammy's comments | Ecency

DiscoverWavesCommunities

chammy

@chammy

53

Finance, Life, Philosophy

Resource Credits

Available

Used

Created

2024-12-22 17:41

RSS Feed

Comments

Blog Posts Comments Communities Wallet

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

Quite a week for open-source AI Especially American open-source Nemotron 3 Ultra is the most important release in quite some time And some really cool RL and fine-tuning work from Harvey

$ 0.000

0

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

As control planes such as Software Factory become more standard, run-rate revenue growth for frontier labs should fall meaningfully while revenues for open models surge — because the nuanced, model-agnostic

$ 0.000

0

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

Most CEOs have no idea that, instead of this nuanced approach, teams are often defaulting to the most expensive models and burning through massive budgets with zero governance, audit ability and control

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

GPT-5.5 Pro only for workloads where its incremental capability demonstrably produces enough business value to justify a 20–40× token premium.

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

When building a company today, the economic frontier looks roughly like: DeepSeek V4 Pro / R1 for high-volume inference. Claude Opus for premium agent workflows where reliability matters.

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

For a company consuming 1 billion input tokens and 1 billion output tokens per month: GPT-5.5 Pro: ~$105,000 Claude Opus 4.8: ~$30,000 DeepSeek V4 Pro: ~$5,220 DeepSeek R1: ~$2,740

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

The pricing gap remains enormous while the capability gap is quite narrow What does this mean in practice?

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

The biggest surprise of 2026 is that the capability gap between the best open-weight/source models and the top closed models has narrowed much faster than the pricing gap.

chammy

2026-06-07 16:06

RE: LeoThread 2026-06-06 13-22

Margins equal opportunity — AI version

chammy

2026-06-07 13:32

RE: LeoThread 2026-06-06 13-22

That represents a much tougher — and lower-return — business than a pure asset-light marketplace, which changes return assumptions That factor appears to have been the primary influence on the stock since

chammy

2026-06-07 13:32

RE: LeoThread 2026-06-06 13-22

Even though the platform is asset-light, after years in the market it has become more levered to the capital investment cycle, hosts’ ability to improve homes and, downstream, hosts’ credit quality

chammy

2026-06-07 13:32

RE: LeoThread 2026-06-06 13-22

Without that, guest quality tends to decline and pricing power follows

chammy

2026-06-07 13:32

RE: LeoThread 2026-06-06 13-22

What few hosts realized when signing up was that after many years of guests their homes would become “well used,” sometimes dilapidated, and require upgrades, updates and general maintenance/TLC.

chammy

2026-06-07 13:32

RE: LeoThread 2026-06-06 13-22

Take Airbnb as an example. Curiosity about why the company stayed range-bound despite strong underlying economics prompted a closer look

chammy

2026-06-07 13:32

RE: LeoThread 2026-06-06 13-22

Capital investment cycles aren't always extremely short like in tech. In many non-tech businesses they can be very long and sneak up on investors, significantly affecting returns

chammy

2026-06-06 12:13

RE: LeoThread 2026-06-05 15-02

Systems are 100 repos, and the risk lives in the seams New multi-repo indexing aims to address this

chammy

2026-06-06 12:13

RE: LeoThread 2026-06-05 15-02

Every platform raced to be the AI agent control plane this week: Snowflake CoCo, Microsoft Agent 365, NVIDIA's faster models Faster agents don't move the real blind spot: AI writes one repo at a time

chammy

2026-06-06 12:13

RE: LeoThread 2026-06-05 15-02

Most importantly, leaders will want INTENT behind AI spending. Vibe coding is dead. Now comes serious ROI-driven analysis, treated like any other major business initiative. That's why Software Factory

chammy

2026-06-06 12:13

RE: LeoThread 2026-06-05 15-02

In the next phase of AI, organizations will want a layer above the models. They will want control over AI spend, the flexibility to choose certain models, tighter control of budgets, and closer ties to

chammy

2026-06-06 12:13

RE: LeoThread 2026-06-05 15-02

Control plane! Control plane! Control plane! This term will be heard a lot going forward. Why?