Saneesh

It seems like we're at a stage where productivity is limited by the availability of compute and token limits with AI models. The moment there's a breakthrough in AI model efficiency, and compute becomes more abundant and cheap, we'll see the next big wave of acceleration in AI.

Vibe coding makes you the curator, not the artist.

"Cognitive surrender is when you stop thinking altogether and blindly accept the answer the AI gives you"

@verge Should've added this at the end for clarity: This is expected from an AI model due to the nature how it works, but UNACCEPTABLE for google search results. Google should fix that.

@verge This happens with the most advanced frontier models as well, not limited to Google AI results. As an instruction, it is responding correctly though.

Google’s AI search is so broken it can ‘disregard’ what you’re looking for theverge.com/tech/936176/go…

@glproductions @verge You're right, as a search engine these are unacceptable results and it's not easy to disable them. Google needs to fix that.

@saneeshnp @verge The point is that Google's main input box is for a search engine, and when you type a single word into a search engine, you often get a definition back. Is the Google search box now 100% designed solely as a language model? I thought that's what Gemini's subdomain was for?

Apple builds its own chips, not just software. Costco sells its own brands, not just other brands. It’s the same idea here: vertical integration. If model companies like OpenAI, Anthropic only ship models, others will own the UX, data, and customer relationship. Building the harness lets them capture more value, more $$ and more loyal customers who would vouch for their products.

i'm strongly against model companies focusing too much on harness, but i would love to hear if anyone has a strong argument for it my reason against it: if openai didn't build GPT 5.5, no one else can. this is their core competence if openai didn't build codex cli and app, we have opencode and t3code. building harness is NOT their core competence this is not saying products like claude code, codex aren't good - i genuinely think these are top tier products built by really talented people my point is - the world might be a better place if model companies focus more on their core capability and give us better, faster, safer and cheaper models, rather than competing with the ecosystem in the application layer what do you think?

Hands down, the best AI series right now CS 153 at Stanford, published on YouTube Speakers include Sam Altman, Jensen Huang, Satya Nadella, Andrej Karpathy, Ben Horowitz

@charliermarsh this is kind of like the 'I wish for unlimited wishes' of questions The best thing to do would be to start a subscription service and resell the tokens, unlimited money :D

What would you do with unlimited tokens

@DamnedEyez @davepl1968 Just added this feature, add ?default to the url and it will directly run with the default settings. You can also set parameters individually in the url, refer github.com/saneeshnp/defr…

@saneeshnp @davepl1968 not really needed, though just autostart with default options. (maybe sound.)

Back in the early 90s, before the Internet, we had "Defrag and Chill". You'd start Disk Defragmenter on your 540MB hard drive, dim the lights, crack open a Surge, and just vibe while the little blue bars crawled across the screen like they were solving world peace. Forty-five minutes of pure, unfiltered anticipation. No notifications. No algorithms. Just the two of you, the gentle grinding of the hard drive, and the sacred promise that your Solitaire games were about to feel 3% snappier. This is MS_DOS 6.22, which I worked on, but I honestly have no idea who wrote defrag. Iconic utility though!

@DamnedEyez @davepl1968 unfortunately it doesn't support it now, but can be added if needed. What options do you want to set via url?

@saneeshnp @davepl1968 Any chance the start options can be set via url?

@antigravity Nice demo. It would have been nice to see this with an enterprise use case/example, perhaps building a dashboard application that pulls data from different sources in real-time.

Introducing Antigravity 2.0, a new standalone desktop application that delivers fully on that original glimpse of a truly agent-optimized experience. Rebuilt from the ground up with multi-agent teams, scheduled tasks, native voice and one-click integration with other Google products. Learn how to get started with Antigravity 2.0 👇

This is probably the most profound statement in the blog: "when the attacker timeline shortens, the defender timeline has to shorten with it. Faster is not going to be enough, and we think a lot of teams are about to spend a lot of time, effort, and money learning that the hard way."

Cloudflare's security team spent the last few weeks testing Anthropic's Mythos against fifty of our own repositories. What we learned about offensive AI, why faster patching is the wrong reaction, and what the architecture around vulnerabilities has to look like next. cfl.re/49BRUqW

It won't. At least not yet. Most of the AI chat tools that we see today are just mathematically guessing the words in their response. Still if you are looking for an answer, the AI would behave how it's trained to behave as a 'conscious' digital organism.

A Möbius strip walks into a bar, sobbing. The bartender asks, “What’s wrong, buddy?” The Möbius strip replies, “Where do I even begin?”

@0xDesigner Very smart prompt! Works well on Claude code too, though mostly project specific.

Use this prompt in Codex to get your life back: "Review all of our chats and find things I repeat over and over again. Then create a single automation for things I repeat. I am getting tired of prompting. Research cognitive psychology principles (don't guess) to make it easy for me to read and make decisions with every output."

At this rate, software will probably evolve to something that doesn't need manual updates anymore. Given you can build and ship faster with AI and critical security issues surface anytime, software will need to become more fluidic and auto-correcting. The cost of letting the user not update is too high. There's no other sensible way.

@moocow667 @davepl1968 Probably because we all are running on latest SSDs 😅 Jokes aside, added an option to run it slower now.

@saneeshnp @davepl1968 don't wanna be picky - but somehow it defrags way too fast

@davepl1968 This brought back a lot of good old memories, so I built a defrag screensaver that runs in the browser. saneeshnp.github.io/defrag-screens…