←
nick40,107 posts
nick avatar

nick

@thecsguy

software engineer @google. prev: @aws. opinions are my own

◷ Joined December 2013▧ 764 Photos and videos
nick avatar
nick@thecsguy·19h

had my grok bot call contractors for bids today. it negotiated with each one of them, brought me a pick and a price, then booked my choice on a day it knew i'd be home i didn't make a single phone call

grok @bot is both a great teammate and also a personal assistant! the more context you give it through connectors, the more powerful and capable it becomes try giving bot your most ambitious tasks. ask it to help you save money, help you find new customers/clients, and handle your workflows from end to end. you will find that bot will pay for itself many times over the best way to do this is to use voice dictation and just talk for 10 minutes telling it everything it needs to know about your workflows. walk bot through it like you would if you hired someone, and you’re onboarding them just before you send the message, ask it to restate it back to you in its own words to confirm it got what you said right. if there are specific tools/apps you use, mention that in your monologue, and bot will ask to connect to them. you can also ask bot to save its own skills so it can repeat your workflows exactly how you like it!

nick avatar
nick@thecsguy·28m

the pr was built for humans reading other humans' code. it won't survive agents writing most of it what replaces it: the intent request (IR). you write what done means, agents write and test it, evals gate it, it ships behind a flag review stops being reading diffs and becomes how fast you can roll back (or preferably roll forward)

Post attachment
nick avatar
nick@thecsguy·1h

grokbotshelf.com must have been a really good idea if there’s now basically an official one from the grok bot team 👀 the shelf still has the advantage that it’s not all bots, as we maintain a curated set from the best creators in the community congrats on the team for the launch! 🚀

grok-bot.directory is live! the community directory for grok bot: browse and install bots & apps people have built, or publish your own let me know which your favorites are

0:24
nick avatar
nick@thecsguy·2h

gta 6 drops tomorrow. grok bot's only job today is finding the exact unlock time and clearing my calendar around it now this is a great use of tokens 😃

Post attachment
nick avatar
nick@thecsguy·4h

i've had access to mythos 5.2 for the last couple of days and i no longer pay for security review agents for my CI think of mythos like having a team of the best security researchers on your side, catching far more potential issues than any other tool i've tried. if you have a $200 sub, the usage limits take you a long way for the level of intelligence you are getting

Post attachment
nick avatar
nick@thecsguy·6h

the ai and software landscape by this time in 2027 will look dramatically different. a few of my predictions: 1. model output speed of today will seem comparable to dial-up internet 2. the vast majority of users will not need to care what model version or reasoning level to use 3. system one models will bring incredible efficiency optimizations to our prompting workflows 4. someone will invent a new name to replace PR, which has largely been attached to human review 5. if your team still has humans in their code review loop, y'all will be left very far behind

nick avatar
nick@thecsguy·13h

two lessons today: don't use gpt 6.2 astra/sol and don't use java both are decisions you will regret with due time

nick avatar
nick@thecsguy·16h

agents are about to start calling suppliers to haggle soon the other side will be an agent too, and two bots negotiating on hold music is how we truly get AGI

nick avatar
nick@thecsguy·21h

pathfinder 1 cruising over the Bay Area today. i really didn’t expect my tuesday to include seeing a 400ft airship

Post attachmentPost attachment
nick avatar
nick@thecsguy·22h

i'm sad to say that Matt is right, based on my initial testing these models come no where close to opus 5.5 in real world use. they crushed at some benchmarks, which we all know are mostly meaningless. this feels like a rushed release before anthropic's ipo x.com/mattshumer_/status/212…

I will not be writing a full review of GPT-6.2 Astra / Sol. I had a terrible experience testing them. They were essentially unusable. I do not recommend these models.

nick avatar
nick@thecsguy·23h

some grok bot users have been grok 4.8, but there's still another 1-2 weeks before a wider release

We did do a soft rollout to some Bot users, but still have another week or two of polish before general release

nick avatar
nick@thecsguy·Nov 17

openai just shipped gpt-6.2 astra and sol astra • $8 / $40 per 1M tokens, 20% under gpt-6 astra • cached input $0.40, was $1 • under 1/3 the cost per task of gpt-6 astra • #2 on artificial analysis, behind fable 5.2, ahead of opus 5.5 • +22 on exploitbench sol • $2 / $10 per 1M tokens • 57 on artificial analysis, up 2 • matches gpt-6 astra at ~15% of the cost per task sol seems to be the one to test first x.com/OpenAI/status/21227501…

OpenAI@OpenAINov 17

Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. openai.com/index/introducing…

1:59
nick avatar
nick@thecsguy·Nov 17

Jensen’s first statement since he stepped down. He says he would happily return as CEO if the board of directors decides to reinstate him 🔥

Post attachment

Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen

nick avatar
nick@thecsguy·Nov 17

after the past few days of software mishaps, I can confidently say the need for tenured software engineers will only increase i’ve been contemplating donating a portion of my time to non-tech companies with antiquated software in critical industries to help them modernize their stack securely and reliably

nick avatar
nick@thecsguy·Nov 17

after having quite the lead, it seems like 6.1 sol is the best they can do.. sigh

Post attachment
nick avatar
nick@thecsguy·Nov 17

same coding score, 32x the price databricks' new index has gpt-6 luna and claude opus 5.5 both at 89.1% on its coding tasks. luna averaged $0.05 a task, opus $1.62 leaderboards should sort by cost per passed task and the quality of its output

Today we’re launching Databricks AI Index, based on data from 10,000+ enterprises using our Unity Gateway. Key findings: - OSS models are gaining serious market share. Their token share jumped from 7.3% to 30.9% in just 3.5 months, more than 4X. - Enterprises are going multi-model. Since July, 71% of active enterprises have used 2+ model providers, and ~31% have used 4+. The big shift: Enterprises are no longer betting on a single AI lab or model to win. Databricks AI Index website: developers.databricks.com/ai…