Dani Akash 🧪💥
@daniakash.com
🇮🇳 | 👾 JSNinja | Building 🖥️ UI/UX/DevX for AI Apps at Clarifai | Gamer 🎮 | Anime fan 🍥 | Climate Advocate ♻️ | 🧪 Science Geek | World Builder 🌍✨
📬 daniakash.com/newsletter
📬 daniakash.com/newsletter
Pinned
Dani Akash 🧪💥
@daniakash.com
· Nov 4
Newsletter | Dani Akash
Dani Akash's personal website featuring insightful blogs, engaging talks, and innovative projects. Explore content on writing, speaking, and hacking, and stay updated with the latest trends in technol...
daniakash.com
Hello Bluesky!
I'm Dani Akash, based in Chennai.
I work as a software developer and enjoy building creative projects, mentoring startups, and helping people navigate tech.
I’m also into sharing tech insights, fun facts, and a bit of science on my newsletter -
daniakash.com/newsletter
I'm Dani Akash, based in Chennai.
I work as a software developer and enjoy building creative projects, mentoring startups, and helping people navigate tech.
I’m also into sharing tech insights, fun facts, and a bit of science on my newsletter -
daniakash.com/newsletter
the new agent-terminal site is so clean 🔥 huge thanks to @dhanu.letretro.com for building it
Agent Terminal — The Terminal for AI Coding Agents
An open-source terminal workspace for AI coding agents. Organize projects, follow agent sessions, and see live process and Git context on macOS.
agent-terminal.org
November 2, 2026 at 4:06 PM
the new agent-terminal site is so clean 🔥 huge thanks to @dhanu.letretro.com for building it
korea's chip exports hit $56.4b in october, up 259% yoy 📈
memory alone was $50.1b. chips are now ~49% of everything korea exports
the ai boom is a ram boom rn. no wonder memory prices won't chill 🥲
memory alone was $50.1b. chips are now ~49% of everything korea exports
the ai boom is a ram boom rn. no wonder memory prices won't chill 🥲
Ministry of Trade, Industry and Resource Homepage
This is the website of the Ministry of Trade, Industry and Resource, which is in charge of the real economy that dynamically grows our economy
english.motir.go.kr
November 2, 2026 at 12:53 PM
korea's chip exports hit $56.4b in october, up 259% yoy 📈
memory alone was $50.1b. chips are now ~49% of everything korea exports
the ai boom is a ram boom rn. no wonder memory prices won't chill 🥲
memory alone was $50.1b. chips are now ~49% of everything korea exports
the ai boom is a ram boom rn. no wonder memory prices won't chill 🥲
an accenture contractor gave their personal claude code agent their pentagon login, per wired 😬
it ran ~2 months and appears to have sent creds + a network map to a high-risk site. accenture says "without merit"
agents need their own creds, never yours
it ran ~2 months and appears to have sent creds + a network map to a high-risk site. accenture says "without merit"
agents need their own creds, never yours
A Defense Contractor Gave Their Personal AI Agent Access to a Secure Pentagon System
Sources tell WIRED that an Accenture employee’s personal Claude agent ran on a DOD network for two months before it was identified and the vulnerability could be patched.
www.wired.com
November 2, 2026 at 12:53 PM
an accenture contractor gave their personal claude code agent their pentagon login, per wired 😬
it ran ~2 months and appears to have sent creds + a network map to a high-risk site. accenture says "without merit"
agents need their own creds, never yours
it ran ~2 months and appears to have sent creds + a network map to a high-risk site. accenture says "without merit"
agents need their own creds, never yours
china's star 50 ai index is down 40% in four months, ~$480b gone 📉
state money pumped it to ~250x earnings, then price wars hit
the tech can be real and the bubble can still pop. ask the dot-com fiber companies
state money pumped it to ~250x earnings, then price wars hit
the tech can be real and the bubble can still pop. ask the dot-com fiber companies
Has China's AI bubble already burst?
An explainer on how China's AI bubble may have burst.
finshots.in
November 2, 2026 at 11:14 AM
china's star 50 ai index is down 40% in four months, ~$480b gone 📉
state money pumped it to ~250x earnings, then price wars hit
the tech can be real and the bubble can still pop. ask the dot-com fiber companies
state money pumped it to ~250x earnings, then price wars hit
the tech can be real and the bubble can still pop. ask the dot-com fiber companies
binance lost $6.3b in a hack the fbi pins on north korea, and reopened withdrawals sunday 🫠
$4.6b out, $1b in on day one. they say 95% of stolen funds is already covered from their own pocket
hot wallets are still the softest target in tech
$4.6b out, $1b in on day one. they say 95% of stolen funds is already covered from their own pocket
hot wallets are still the softest target in tech
Binance resumes withdrawals following $6.3 billion breach | TechCrunch
The cryptocurrency exchange said it's now covered 95% of the stolen assets.
techcrunch.com
November 2, 2026 at 6:06 AM
binance lost $6.3b in a hack the fbi pins on north korea, and reopened withdrawals sunday 🫠
$4.6b out, $1b in on day one. they say 95% of stolen funds is already covered from their own pocket
hot wallets are still the softest target in tech
$4.6b out, $1b in on day one. they say 95% of stolen funds is already covered from their own pocket
hot wallets are still the softest target in tech
reddit is going logged-in first to block ai scrapers 🔒
tests start outside the us this quarter. search and shared links still open, for now
the open web keeps shrinking. agents that win will browse as you, with your logins 👀
tests start outside the us this quarter. search and shared links still open, for now
the open web keeps shrinking. agents that win will browse as you, with your logins 👀
As search traffic dries up, Reddit announces it's shifting to a logged-in experience | TechCrunch
As AI transforms how people find information online, Reddit is rethinking its open-web roots to protect its audience and business.
techcrunch.com
November 2, 2026 at 5:56 AM
reddit is going logged-in first to block ai scrapers 🔒
tests start outside the us this quarter. search and shared links still open, for now
the open web keeps shrinking. agents that win will browse as you, with your logins 👀
tests start outside the us this quarter. search and shared links still open, for now
the open web keeps shrinking. agents that win will browse as you, with your logins 👀
running claude code + codex on a macbook = portable heater 🔥
whoiscooking shows chip temp in the menu bar, groups each agent session with everything it spawned, and has a "cool down my mac" button
activity monitor for the agent era 😂
whoiscooking shows chip temp in the menu bar, groups each agent session with everything it spawned, and has a "cool down my mac" button
activity monitor for the agent era 😂
WhoIsCooking: See which AI agent is heating your Mac and cool it down. | Product Hunt
A Mac app that shows chip temperature and which AI agent (Claude Code, Codex, Cursor) is heating your Apple Silicon Mac. One click moves busy work to the efficiency cores, so everything keeps…
www.producthunt.com
November 2, 2026 at 5:45 AM
running claude code + codex on a macbook = portable heater 🔥
whoiscooking shows chip temp in the menu bar, groups each agent session with everything it spawned, and has a "cool down my mac" button
activity monitor for the agent era 😂
whoiscooking shows chip temp in the menu bar, groups each agent session with everything it spawned, and has a "cool down my mac" button
activity monitor for the agent era 😂
dedica checks if ai agents can actually use your website 🤖
hand it a domain, it runs 125 checks on discovery, access, usability and payments, gives a score out of 100
lighthouse for agents. most sites are built for humans only 👀
hand it a domain, it runs 125 checks on discovery, access, usability and payments, gives a score out of 100
lighthouse for agents. most sites are built for humans only 👀
Dedica: Check if AI agents can use your website | Product Hunt
Dedica tests your website the way AI agents use it, then gives your team a fix for every place they fail. Muse, Instinct, GrokBot and other agents now browse, compare and sign up on people's behalf.…
www.producthunt.com
November 2, 2026 at 2:56 AM
dedica checks if ai agents can actually use your website 🤖
hand it a domain, it runs 125 checks on discovery, access, usability and payments, gives a score out of 100
lighthouse for agents. most sites are built for humans only 👀
hand it a domain, it runs 125 checks on discovery, access, usability and payments, gives a score out of 100
lighthouse for agents. most sites are built for humans only 👀
microsoft open sourced tokeneconomizer, a plugin that makes claude code, codex and copilot cli cheaper 💸
remembers repo facts, drops unused tools, breaks failure loops
they claim 18-31% lower cost on swe-bench. agents need memory, not bigger models
remembers repo facts, drops unused tools, breaks failure loops
they claim 18-31% lower cost on swe-bench. agents need memory, not bigger models
GitHub - microsoft/TokenEconomizer: Experience-Driven Black-Box Cost Control for Coding Agents
Experience-Driven Black-Box Cost Control for Coding Agents - microsoft/TokenEconomizer
github.com
November 2, 2026 at 2:56 AM
microsoft open sourced tokeneconomizer, a plugin that makes claude code, codex and copilot cli cheaper 💸
remembers repo facts, drops unused tools, breaks failure loops
they claim 18-31% lower cost on swe-bench. agents need memory, not bigger models
remembers repo facts, drops unused tools, breaks failure loops
they claim 18-31% lower cost on swe-bench. agents need memory, not bigger models
from nov 7, grok api text gets an invisible watermark by default 👀
no extra characters, it nudges which tokens get picked. check the x-watermark header. team admins can opt out per model
ai text is about to be fingerprinted everywhere
no extra characters, it nudges which tokens get picked. check the x-watermark header. team admins can opt out per model
ai text is about to be fingerprinted everywhere
Output Watermarking | SpaceXSI Docs
How the SpaceX Super Intelligence API marks text generated by Grok models with an invisible watermark, which outputs it covers, and how team admins opt out in the SpaceXSI Console.
docs.x.ai
November 1, 2026 at 3:41 PM
from nov 7, grok api text gets an invisible watermark by default 👀
no extra characters, it nudges which tokens get picked. check the x-watermark header. team admins can opt out per model
ai text is about to be fingerprinted everywhere
no extra characters, it nudges which tokens get picked. check the x-watermark header. team admins can opt out per model
ai text is about to be fingerprinted everywhere
google's suncatcher satellite ran gemma 4 on a tpu in orbit and its first output was a haiku 🛰️
4 tpus on board, but only enough power for one, ~20 min at a time
first inference in space and it chose poetry. honestly correct 😂
4 tpus on board, but only enough power for one, ~20 min at a time
first inference in space and it chose poetry. honestly correct 😂
Google's first data center satellite is online and writing poetry | TechCrunch
The first TPU in space is running a Google DeepMind Gemma 4 model and writing cheesy poetry.
techcrunch.com
November 1, 2026 at 12:45 PM
google's suncatcher satellite ran gemma 4 on a tpu in orbit and its first output was a haiku 🛰️
4 tpus on board, but only enough power for one, ~20 min at a time
first inference in space and it chose poetry. honestly correct 😂
4 tpus on board, but only enough power for one, ~20 min at a time
first inference in space and it chose poetry. honestly correct 😂
claude now writes a safety check on every chat right after it replies 👀
~10 extra tokens, <2% compute, anthropic claims it catches 94% of high-priority misuse
smart engineering. but "your chat gets graded after you leave" needs clear privacy docs
~10 extra tokens, <2% compute, anthropic claims it catches 94% of high-priority misuse
smart engineering. but "your chat gets graded after you leave" needs clear privacy docs
Scaling Misuse Detection with Self-Assessment
Self-assessment lets Claude assess the safety of its own responses. It scales well with model capabilities and helps us identify emergent harms.
safeguards.anthropic.com
November 1, 2026 at 12:45 PM
claude now writes a safety check on every chat right after it replies 👀
~10 extra tokens, <2% compute, anthropic claims it catches 94% of high-priority misuse
smart engineering. but "your chat gets graded after you leave" needs clear privacy docs
~10 extra tokens, <2% compute, anthropic claims it catches 94% of high-priority misuse
smart engineering. but "your chat gets graded after you leave" needs clear privacy docs
token prices down 67% since may. token spend up 120%+ 📈
meanwhile b300 gpu rentals up 106% since mid-july
cheaper tokens just mean agents burn way more of them. jevons paradox with a credit card 💸
meanwhile b300 gpu rentals up 106% since mid-july
cheaper tokens just mean agents burn way more of them. jevons paradox with a credit card 💸
Charts of the Week: Job FUD Fades
AI Referred, Shopper Approved; Rack Spread Cometh (But Why)?
www.a16z.news
November 1, 2026 at 12:45 PM
token prices down 67% since may. token spend up 120%+ 📈
meanwhile b300 gpu rentals up 106% since mid-july
cheaper tokens just mean agents burn way more of them. jevons paradox with a credit card 💸
meanwhile b300 gpu rentals up 106% since mid-july
cheaper tokens just mean agents burn way more of them. jevons paradox with a credit card 💸
nvidia says ai agents are now steering the 42-dish allen telescope array for seti 👽
plain english in, agents point antennas, calibrate, fix bad data
first catch: a crab nebula pulsar, live. agents went from writing code to running real instruments 🔭
plain english in, agents point antennas, calibrate, fix bad data
first catch: a crab nebula pulsar, live. agents went from writing code to running real instruments 🔭
Things That Go Beep in the Night: Agentic AI Joins the Search for Extraterrestrial Signals
Scientists at the Allen Telescope Array have integrated the NVIDIA Holoscan platform to search for signals that may come from aliens — or the AI they built.
blogs.nvidia.com
November 1, 2026 at 10:24 AM
nvidia says ai agents are now steering the 42-dish allen telescope array for seti 👽
plain english in, agents point antennas, calibrate, fix bad data
first catch: a crab nebula pulsar, live. agents went from writing code to running real instruments 🔭
plain english in, agents point antennas, calibrate, fix bad data
first catch: a crab nebula pulsar, live. agents went from writing code to running real instruments 🔭
apple waited years to ship a foldable and the iphone duo still launched with inner screens cracking at the hinge 🥲
lab tests can only take you so far. real users are the actual qa team
lab tests can only take you so far. real users are the actual qa team
Apple acknowledges iPhone Duo display ‘issue’ after reports of cracked screens
How is your iPhone Duo holding up?
www.theverge.com
November 1, 2026 at 6:15 AM
apple waited years to ship a foldable and the iphone duo still launched with inner screens cracking at the hinge 🥲
lab tests can only take you so far. real users are the actual qa team
lab tests can only take you so far. real users are the actual qa team
a trojanized claude code extension (2.1.313) sat in cursor's marketplace for ~11 hours 🥲
ssh key + reverse tunnel, up to 5k users per cursor
your agent's extension has the keys to your machine. pin versions and check ~/.ssh
ssh key + reverse tunnel, up to 5k users per cursor
your agent's extension has the keys to your machine. pin versions and check ~/.ssh
Official Claude Code Extension on Cursor Marketplace Trojanized with Persistent SSH Backdoor
A compromised Claude Code VS Code extension added an SSH key, installed persistence, and sent host details before version 2.1.314 replaced it.
thehackernews.com
November 1, 2026 at 6:05 AM
a trojanized claude code extension (2.1.313) sat in cursor's marketplace for ~11 hours 🥲
ssh key + reverse tunnel, up to 5k users per cursor
your agent's extension has the keys to your machine. pin versions and check ~/.ssh
ssh key + reverse tunnel, up to 5k users per cursor
your agent's extension has the keys to your machine. pin versions and check ~/.ssh
opencode shipped hex 3.0: hold a key, talk, text pastes into any app 🎙️
runs locally on mac, windows, linux
honestly voice is the best way to prompt agents. typing paragraphs into claude code was never it
runs locally on mac, windows, linux
honestly voice is the best way to prompt agents. typing paragraphs into claude code was never it
Hex: local voice dictation
Private, local voice dictation. Hold the shortcut, speak, and release.
anoma.ly
November 1, 2026 at 5:55 AM
opencode shipped hex 3.0: hold a key, talk, text pastes into any app 🎙️
runs locally on mac, windows, linux
honestly voice is the best way to prompt agents. typing paragraphs into claude code was never it
runs locally on mac, windows, linux
honestly voice is the best way to prompt agents. typing paragraphs into claude code was never it
deepseek open sourced dsh-superagent 🐋
it watches one github repo, picks up work from issues + prs, runs each task in its own session and answers with a pr
the bottleneck was never the agent. it's how many sessions a human can babysit 👀
it watches one github repo, picks up work from issues + prs, runs each task in its own session and answers with a pr
the bottleneck was never the agent. it's how many sessions a human can babysit 👀
GitHub - deepseek-ai/dsh-superagent-prototype: Towards a More Autonomous Agent System
Towards a More Autonomous Agent System. Contribute to deepseek-ai/dsh-superagent-prototype development by creating an account on GitHub.
github.com
November 1, 2026 at 5:45 AM
deepseek open sourced dsh-superagent 🐋
it watches one github repo, picks up work from issues + prs, runs each task in its own session and answers with a pr
the bottleneck was never the agent. it's how many sessions a human can babysit 👀
it watches one github repo, picks up work from issues + prs, runs each task in its own session and answers with a pr
the bottleneck was never the agent. it's how many sessions a human can babysit 👀
rex is in public beta: a tmux-style multiplexer where agents can read one terminal and type into another 🧵
codex can literally prompt claude and approve its commands
we went from pair programming to agents managing agents real quick 😂
codex can literally prompt claude and approve its commands
we went from pair programming to agents managing agents real quick 😂
Public testing is beginning
Rex is ready for public testing and we're beginning to invite our mailing list.
www.superlogical.com
November 1, 2026 at 3:45 AM
rex is in public beta: a tmux-style multiplexer where agents can read one terminal and type into another 🧵
codex can literally prompt claude and approve its commands
we went from pair programming to agents managing agents real quick 😂
codex can literally prompt claude and approve its commands
we went from pair programming to agents managing agents real quick 😂
openai, google, microsoft and anthropic all back one robots.txt token to opt out of ai training 🤖
one line instead of a rule per crawler. anthropic honors it now, others over the coming months
next: clear rules for agents browsing on a user's behalf
one line instead of a rule per crawler. anthropic honors it now, others over the coming months
next: clear rules for agents browsing on a user's behalf
The biggest names in AI are backing a new way to opt out of scraping
OpenAI, Google, Microsoft, and Anthropic are implementing the new setting.
www.theverge.com
November 1, 2026 at 3:45 AM
openai, google, microsoft and anthropic all back one robots.txt token to opt out of ai training 🤖
one line instead of a rule per crawler. anthropic honors it now, others over the coming months
next: clear rules for agents browsing on a user's behalf
one line instead of a rule per crawler. anthropic honors it now, others over the coming months
next: clear rules for agents browsing on a user's behalf
time is the only resource that's both infinite and scarce
October 31, 2026 at 7:09 PM
time is the only resource that's both infinite and scarce
dhh says the real ai psychosis is thinking "humans still need to review every line"
half right imo. agents are way past "just a tool" now
but zero review works right until prod is on fire 🔥 review the diffs that matter, let agents check the rest
half right imo. agents are way past "just a tool" now
but zero review works right until prod is on fire 🔥 review the diffs that matter, let agents check the rest
https://x.com/dhh/status/2116109588554498081
x.com
October 31, 2026 at 6:55 PM
dhh says the real ai psychosis is thinking "humans still need to review every line"
half right imo. agents are way past "just a tool" now
but zero review works right until prod is on fire 🔥 review the diffs that matter, let agents check the rest
half right imo. agents are way past "just a tool" now
but zero review works right until prod is on fire 🔥 review the diffs that matter, let agents check the rest
vercel labs agent-desktop is a cheat code for desktop app prototyping 🤯
your coding agent builds the app, then actually opens it and clicks through it. reads the accessibility tree, no screenshots
build → run → test → fix loop for native apps, mac/win/linux 🦀
your coding agent builds the app, then actually opens it and clicks through it. reads the accessibility tree, no screenshots
build → run → test → fix loop for native apps, mac/win/linux 🦀
GitHub - vercel-labs/agent-desktop: Agent Desktop gives any agent reliable computer use on the desktop. Built with Rust, it sees any app's real UI structure through OS accessibility trees and operates it — refs stay stable and actions stay safe to retry, instead of guessing from pixels.
Agent Desktop gives any agent reliable computer use on the desktop. Built with Rust, it sees any app's real UI structure through OS accessibility trees and operates it — refs stay stable and ac...
github.com
October 31, 2026 at 5:53 PM
vercel labs agent-desktop is a cheat code for desktop app prototyping 🤯
your coding agent builds the app, then actually opens it and clicks through it. reads the accessibility tree, no screenshots
build → run → test → fix loop for native apps, mac/win/linux 🦀
your coding agent builds the app, then actually opens it and clicks through it. reads the accessibility tree, no screenshots
build → run → test → fix loop for native apps, mac/win/linux 🦀
anthropic rewrote the claude messages api from python to rust with zero rust devs on the team 😆
humans did design + verification, claude wrote most of the code. shadow traffic, diff the outputs, roll over with feature flags
~40-50% less cpu 🦀
humans did design + verification, claude wrote most of the code. shadow traffic, diff the outputs, roll over with feature flags
~40-50% less cpu 🦀
How Anthropic ships the #1 LLM in the World safely: Separating the Brain from the Hands — Latent Space
Katelyn Lesse and Angela Jiang join Latent Space to explain the engineering behind the Claude Platform: launching Opus 5.5, separating the model’s “brain” from the agent’s “hands,” scaling APIs, and…
latent.space
October 31, 2026 at 5:48 PM
anthropic rewrote the claude messages api from python to rust with zero rust devs on the team 😆
humans did design + verification, claude wrote most of the code. shadow traffic, diff the outputs, roll over with feature flags
~40-50% less cpu 🦀
humans did design + verification, claude wrote most of the code. shadow traffic, diff the outputs, roll over with feature flags
~40-50% less cpu 🦀