HNHacker News
TopNewBestAskShowJobs

tesnorindian

61 karma · joined June 29, 2026

An average Indian engineer who would like to change this world
submissionscomments
tesnorindian··on Samsung's Processing-in-Memory (PIM)
In memory compute has been in talks since LLMs took up. I remember few flocks were trying to get RISC V cores in the memory like these papers

https://arxiv.org/abs/2602.01827

tesnorindian··on Nvidia agrees to acquire Hugging Face for $13B
Now what will happen to the llama.cpp performance on non Nvidia GPUs and especially on arm64 CPU chips after this acquisition? Will Apple M* chips and cloud arm64 chips (Graviton, Cobalt, Axion) become second class?

Just now I got my hands wet with local LLMs before this was announced. Can somebody from llama.cpp/ggml.ai confirm this?

tesnorindian··on Actually Queryable Executables
It took me several minutes to understand how this even works. Brilliant hack of using binfmt_misc rules under hood which is like a shebang for scripts but a custom interpreter. The idea of moving the ELF byte code to the segments table and executing the web server from there is next level crazy.

Good for deployment but an accidental deletion of the binary can cause loss of both data and code. Good approach for AI harness and agents though.

tesnorindian··on OpenAI Jalapeño: Better than Nvidia Blackwell
Our brain is more like a MoE model activating only a few neurons for specific activities making it more efficient unlike a dense model activating all the params.

Also brain produces quality tokens @ 3.3 tps instead of fast generating hallucinated tokens by certain models. Thus MTP can produce low quality tokens at 2x speed.

Patience pays.

tesnorindian··on Nitter and XCancel receive cease and desist notices
Xcancel was my only window to X.com which I prefer for our hyperlocal news. With the mainstream media not covering hyperlocal news I am now denied a freedom access to our hyperlocal content. It is a shame that we are denied freedom access to content in this AI slop era.
tesnorindian··on Headlong: A microharness for persistent agents
+1
tesnorindian··on Headlong: A microharness for persistent agents
Looks very interesting. Best suited with Local LLMs with unlimited tokens.
tesnorindian··on Ask HN: What are you working on? (August 2026)
smol code has RCE risk granting model unrestricted access to local environment.
tesnorindian··on Ask HN: What are you working on? (August 2026)
Making legacy 20+ years Java Swing apps run inside a Browser using WASM and JS glue code with CheerpJ/OpenJDK runtime. WASM has a bright future for running native apps. Both WASM and JVM are stack machines. WASM beats on startup speed and memory footprint. But performs bad on compute. Since WASM lacks networking sockets, JDBC is broken.
tesnorindian··on AMD acquires Taalas to boost inference performance by etching models in silicon
We also have ReRAM (Analog Computing), which also holds a promising future given its efficiency and low power. Though ReRAM of larger size is still a research area.
tesnorindian··on Atlassian Rovo Exfiltrates Data, Bypassing Controls
This is the reason I show my frustration in my Jira tickets so that let Rovo learn how frustrated humans have became with AI slop.
tesnorindian··on Qwen3.8-Max: A New Bar for Coding and Cowork
AWS Bedrock still lacks the support for the latest open weights models like GLM 5.2, DSV4 Flash 0731, Kimi K3 etc. Even they only support Qwen 3 which is a very old model. Any specific reasons they are reluctant to support open weights models?
tesnorindian··on Read this before you buy that TV streaming stick
LibreElec also supports ARM builds than can run on SBC like Raspberry Pi. While CoreElec is exclusively for Amlogic ARM processors.
tesnorindian··on A California aquifer may have crossed the point of no return
+1 for moving water from wet areas to dry areas. But desalination plants comes at a huge cost.
tesnorindian··on A California aquifer may have crossed the point of no return
Yes, looks like they are acting like a sponge whose efficiency to retain water decreases as time passes.
tesnorindian··on A California aquifer may have crossed the point of no return
Most rivers drain into Oceans while some do drain into inland lakes.
tesnorindian··on A California aquifer may have crossed the point of no return
As the Super El-Nino is at its peak, South India is facing an acute water crisis

https://www.thehindu.com/news/national/cauvery-water-managem...

#SaveWater #SaveElectricity #SaveFuel

Interestingly none of our Govts. or Leaders are talking about conservation yet and the ancient water laws/standards don't take climate change into consideration. For eg. as per BIS IS 1172: 1993 standards an average Indian is entitled to 135 Liters of water per day when the actual supply is not even close to 50 liters per Indian per day. This is level of water crisis India is facing today.

tesnorindian··on Codex Security
Our employer uses it unaware of its links.
tesnorindian··on Codex Security
This bundles with 13 security skills. Hard coded to gpt-5.6-sol unlike Codex. Runs in an isolated sandbox via a cloned home directory.
tesnorindian··on Kimi-K3 Technical Report [pdf]
> Two findings in the Linux kernel illustrate the depth of these results. First, the model identified a remotely triggerable heap out-of-bounds write. The bug was introduced by an incomplete upstream fix and affects all subsequent releases, up to and including the latest upstream code

Hope these Linux Kernel bugs due to some incomplete upstream fix identified by K3 patched? Moonshot did not provide much details on this.

https://www.opensourceforu.com/2026/07/linux-maintainers-bat...

Looks like K3 is good with identifying zero days

https://www.heise.de/en/news/Kimi-K3-Chinese-AI-finds-severa...

tesnorindian··on Exploiting Volvo/Eicher's fleet platform to gain control over all users/vehicles
No wonder a few weeks before few BMS apps based on BT was banned in India as they were misused for remote disabling of e-Rickshaws.

https://timesofindia.indiatimes.com/business/india-business/...

Securing BMS should be the top priority for EVs. Keeping unsecured BT connection in BMS would be the worst thing.

tesnorindian··on Netflix employee fired for sharing personal details in retreat trust exercise
People should reject this mentality and live their life.
tesnorindian··on Netflix employee fired for sharing personal details in retreat trust exercise
+1. I too feel the same thing after 3 years of remote work. Sometimes I even feel that my colleagues are way more dumb than they appeared to look in online.
tesnorindian··on Our position on open-weights models
+1 and this will really be a good move. Open weights models are not going anywhere the focus should now shift to hardware and compute. Compute needs to be optimized for LLMs, currently they are inefficient. Analog computing like ReRAM should get more focus.
tesnorindian··on Our position on open-weights models
AI should not be centralized and should be distributed. There can be no 1 provider to serve all of our needs. We would require both Open weights and Closed weights models for a lot of our use cases.

Open weights models can be leveraged to optimize the cost of Closed weights models. Open weights models can be leverage to defend cyberattacks as HF has shown. Closed weights models can too act as a better cyberattack defender provided separate subscription exists for those.

More efforts are required on LLM distillation for several edge cases. LLM weights should be optimized and compressed to run on edge devices (K3 on Pi3 :). Distillation should be seen as a cost optimization strategy rather than as a competition. You cannot prevent a teacher from teaching to students. If not from teacher A, I will learn from teacher B, you cannot prevent my continuous learning.

tesnorindian··on The New AI Superpowers: Focus and Followthrough
I am going through this burnout cycle since the COVID-19 days. I noticed that the origin of my burnout is not just due to my job, but also several other factors. Though my job constitutes 80% of it. Within that 80% I noticed that doing a monotonous job contribute to 80% and remaining due to orgn. policies, boss and coworkers.

When I first started using AI agents to automate my work tasks I felt liberated for a few months, but after that I again started feeling burnout. There were days when I tried to compete with AI Agents to complete my tasks earlier like zooming in a race car which added to my burnout.

The quest to learn or gain knowledge will alleviate your burnout and the moment you start unlearning your burnout reappears. Thus I stick to HN for most of the day to kill my burnout though it is overwhelming.

tesnorindian··on OverpAId – Fire your CEO. Hire the future
It is plain simple that the board requires a scapegoat.
tesnorindian··on Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
This pelican is the fattest with a long neck I have ever seen.

Does this say anything about the model? I meant the underlying attention/pattern it took for Gemini 3.6 Flash to create this SVG.

tesnorindian··on Petals: Run LLMs at home, BitTorrent-style
https://public.meshllm.cloud/ - This should give the status of their public mesh. On an average I notice around 10 nodes on their public mesh serving various SLMs. You can use meshllm to share your GPUs like SETI@Home. Wondering how security is handled at a random node.
tesnorindian··on OpenAI and Hugging Face address security incident during model evaluation
It is the OAI ExploitGym agents (on GPT 5.6-Sol with guardrails turned off) that escaped the sandbox, found a zero day in HF production dataset and exploited it.
← PreviousPage 3 of 4Next →