SpeakEZ.tech
@SpeakezTech
Followers
104
Following
2K
Media
28
Statuses
642
Deep tech (4 patents) fusing #HPC, #AI & #PQC. Hardware-software fusion for CPU, #GPU, #TPU, #FPGA, #CGRA & neuromorphic in our Fidelity framework. #fsharp
United States
Joined August 2025
🚂💨 #AI's dirty secret: Everyone is driving steam-powered cars pretending they're flying. ⛏️🔥 While #OpenAI, #Microsoft, #AWS and #Google burn megawatts for chatbots, they're continuing to shovel matrix multiplications like coal into billowing furnaces. The Fidelity
speakez.tech
How Fidelity Transforms the AI Performance Equation for Both Current and Emerging Hardware
1
1
3
Checks notes: End of 2025? Yes Number of autonomous "Robotaxis"? Zero Anybody surprised?
271
580
5K
China built a working EUV prototype in less than 5 years and the entire venture capital industry hasn’t produced one successful lithography startup. I don’t say this with like joy or to rub it in. But we need to take a hard look at what we are doing wrong. I don’t think the
184
172
2K
Danish head of government IT (left) hands over the first "microsoft-free" computer to the head of Danish Traffic control, December 2025. The EU is slowly becoming tech independent from *big tech* :)
27
76
759
History rhymes
0
0
0
"[T]here’s a reason this 'tech-friendly dictator' fantasy has been bubbling up in VC circles and startup accelerators for years."
Excellent piece - “Fascism For First Time Founders” https://t.co/XDIbONB55B
0
0
1
We’re seeing up to 14.5 million attempts per hour to exploit the #React2Shell vulnerability across @Cloudflare’s network. This is a very bad exploit. While our WAF is helping protect customers, it’s critical to update your React and Next.js instances as soon as possible.
20
139
1K
2024: 'AI will change everything.' 2025: 'AI will change everything.' 2026: 'AI will change everything.' At some point, something should change.
252
64
1K
There is no AI race. None of what they're building is AI. But it all HAS been a tremendously criminal waste of time, money, and resources. You have unfortunately been fooled.
DeepSeek's ability to access advanced Nvidia chips allows it to stay competitive in the global AI race, as Chinese AI chips are still not good enough for training AI models. More details: https://t.co/m9nayHCF9p
16
884
6K
People are obsessing over longevity while ignoring that medical research funding has been obliterated.
4
3
91
$ORCL bond holders now sit on 9% of unrealized losses on $18 billion of debt issued just this September with the bonds falling to “junk” rating as CDS jump from 1.58 on December 9th to 1.71 December 12, now matching its highest number since the Great Recession. Oracle has made
bloomberg.com
Get caught up.
159
538
2K
Heard from a friend today in the UK who has a political Substack. Age verification is, quote, "killing his business." The Online Safety Act censorship ain't about kids, it's about strangling political viewpoints the Act requires platforms to censor.
34
419
1K
#CancelDataCenters Make Hyperlocal Great Again
Fermi Craters 50% After Losing First Tenant For Its Massive Texas Data Center #MacroEdge
0
0
0
These people used to sue five-year-olds for having Mickey Mouse on their birthday cake
Disney is reportedly investing $1 BILLION in OpenAI, licensing over 200 of their iconic characters for ChatGPT & Sora. ( https://t.co/joJnXQu505)
232
22K
235K
It was never about curing diseases or solving the climate crisis. Never. F*ck these people. This bubble cannot pop soon enough.
OpenAI Sam Altman says the demand for Disney characters for their users has been "off the charts." "It is going to be a wonderful start for what our customers want to do when it comes to putting themselves in a lightsaber fight in Star Wars or making like a Buzz Lightyear custom
38
2K
14K
"We came together as a company" is not the flex you thought was from inside the monopolist. 🙄
We came together as a company to redesign the Gemini API with Interactions API. We support storage, long-running execution, MCP support out of the box. It serves both models and agents. Many teams worked together to make this possible! https://t.co/694KHIh826
0
0
1
When Meta trains it models on 80+ TB of pirated books from LibGen and other platforms, it's called 'fair use', without them having to pay penalties and / or receive some form of legal punishment, as proceedings are ongoing. When Aaron Swartz downloaded 70 GB of articles from
308
7K
33K