ossa-ma

8 ms·

91.
by ossa-ma 9mo ago
A decade of reinforcement and agentic learning was spent playing games (Google Deepmind AlphaGo, AlphaStar, OpenAI Five), including against each other. So what makes it a new frontier?
92.
by ossa-ma 9mo ago
Good post, outlines what a GOOD senior/staff engineer does well and identifying which of these LLMs are missing. > But the model may need to actually do wetwork (e.g., talk to humans) to get all of the information it needs. Even thi
93.
The Silence of the LLaMbs: Getting LLMs to Shut Up(ossa-ma.github.io)
5 points by ossa-ma 9mo ago|0 comments
94.
by ossa-ma 9mo ago
I've adjusted the statistic for you: Adjusted beef consumption: 4.5 million litres of water can be used to produce 300kg of beef -> US (highest beef consumer/capita) consumes 23.3kg of beef , enough to feed ~13 Americans (30 Br
95.
by ossa-ma 9mo ago
And the other stats you ignored? As well as the main point I was making?
96.
by ossa-ma 9mo ago
"The University of Rhode Island based its report on its estimates that producing a medium-length, 1,000-token GPT-5 response can consume up to 40 watt-hours (Wh) of electricity, with an average just over 18.35 Wh, up from 2.12 Wh for G
97.
by ossa-ma 9mo ago
So he's spent $51k on tokens in 3 months to build what exactly? Tools to enable you to spend more money on tokens? Quick math on the environmental impact of this assuming 18.35Wh/1000 tokens: Total energy: 4.73GWh, equivalent of p
98.
by ossa-ma 9mo ago
The beginning perfectly embodies the culture in Silicon Valley and touches on a crucial part that I notice when I visit: the complete lack of self expression or as I would put it ZERO drip. Remove the tech, what does SF contribute to the wo
99.
by ossa-ma 9mo ago
I like to read and review what was captured in .inbox.md before it is committed and synced across my knowledge base. Allows me to catch mistakes, tweak preferences, add context and decide whether something is actually worth pushing. I will
100.
by ossa-ma 9mo ago
There are a quadrillion startups (mem0, langmem, zep, supermemory), open source repos (claude-mem, beads), and tools that do this. My approach is literally just a top-level, local, git version controlled memory system with 3 commands: - &#x
101.
by ossa-ma 9mo ago
Sorry responded to the wrong person!
102.
Nvidia Groq Update: Everyone Gets Rich, Patent Warfare Begins(ossa-ma.github.io)
7 points by ossa-ma 10mo ago|3 comments
103.
by ossa-ma 10mo ago
LMAO you got me
104.
by ossa-ma 10mo ago
Very interesting would love to read about this, do you have a link to the article?
105.
by ossa-ma 10mo ago
You're right but my understanding is that Groq's LPU architecture makes it inference-only in practice. Like Groq's chips only have 230MB of SRAM per chip vs 80GB on an H100, training is memory hungry as you need to hold model
106.
by ossa-ma 10mo ago
Part of it is the irrational feeling of a disgruntled investor, doubt that I'm alone. The other part is he has a track record of dumping on retail then telling them not to buy his next deal once he's already cashed out.
107.
by ossa-ma 10mo ago
No shade but most other coverage will focus on whether this signals an AI bubble. That's missing the story. Nvidia explicitly did NOT acquire Groq. They licensed the IP and hired the talent. This structure dodges CFIUS review (Groq had
108.
Nvidia's $20B antitrust loophole(ossa-ma.github.io)
549 points by ossa-ma 10mo ago|175 comments
109.
by ossa-ma 10mo ago
The bubble take is tired. This was regulatory arbitrage: IP licensing instead of acquisition to dodge CFIUS/antitrust. The $13B premium to avoid years of hold up while enriching Chamath and giving Trump's AI Czar a Christmas prese
110.
Indexing: From Google to Shazam to AI Agents(ossa-ma.github.io)
1 points by ossa-ma 10mo ago|0 comments
111.
by ossa-ma 10mo ago
What does it mean that 70% of employee titles are 'Member of Technical Staff'? I feel this role is becoming less transparent yet more popular in ML.
112.
When not to use Pydantic(ossa-ma.github.io)
4 points by ossa-ma 10mo ago|0 comments
113.
by ossa-ma 10mo ago
This is more or less what private equity does and has been doing for years but they've added AI to the loop. There are companies that do this with AI too, notably Bending Spoons.
114.
by ossa-ma 10mo ago
The core difference I've seen is that FDEs sell a product whereas consultants sell billable hours. The Big 4 incentive is to stay as long as possible, build custom code that only they can maintain, upcharge on maintenance while outsour
115.
by ossa-ma 10mo ago
> Would you sell MSFT because of their involvement in Gaza? Yes. > Sell a broad market index because a company there is doing something 'immoral'? Honestly, this is a lot harder to do. It depends on your definition of immora
116.
FDEs were why I invested in Palantir in 2022 (and sold it all in 2024)(ossa-ma.github.io)
7 points by ossa-ma 10mo ago|6 comments
117.
Crashing an AI Promo Event: What to Ask Before Buying into an AI Agent Platform(ossa-ma.github.io)
1 points by ossa-ma 10mo ago|0 comments
118.
by ossa-ma 10mo ago
The biggest issue with Antigravity is that it completely freezes everything: the IDE, the terminals, debugger, absolutely everything completely blocking your workflow for minutes when running multiple agents, or even a single agent processi
119.
by ossa-ma 11mo ago
Kinda useful, especially tip 15 and tip 26. There needs to be a lot more focus on the observability and showing users what is happening underneath the hood (especially wrt costs and context management for non-power users). A useful feature
120.
by ossa-ma 11mo ago
AGI? In a single prompt? I guess we should tell thousands of AI researchers to stop what they're doing right now since you're a single prompt away from solving the problem??

More ›