o.o

i'm building a fuel finder app on a borrowed 2015 macbook air with 4gb ram. it crashes. it restarts randomly. cursor takes a full minute to load. but i'm two tweaks away from deploying on the play store. your setup is not your ceiling.

@gdb the hard tasks were always the point. easy ones just needed better prompts

gpt-5.5 great for hard tasks like writing GPU kernels

@googleafrica @MoCDTI the best part about this isn't the certificate. it's 100k people in ghana about to realize they can actually do this stuff.

🇬🇭 Big news from Ghana! Applications are open for 100k Google Career Certificate scholarships!🎉. Together with the 1 Million Coders (OMC) program & @MoCDTI , we're equipping youth with job-ready skills in Cybersecurity, AI, UX Design & more. Build your tech career with no prior experience. All the details ▶️ goo.gle/100kCareerScho… #GoogleCareerCerts

@svpino this is the difference between people who use claude and people who actually get work done with claude. most people don't even know /rewind exists.

"Summarize from here" in Claude Code is awesome: Here is how it works: 1. Hit Esc+Esc or type /rewind. This will open the checkpoint menu with a scrollable list of all checkpoints Claude has created during the session. 2. Pick the checkpoint you want to summarize from. Everything before this checkpoint stays intact, and everything after it will get compressed. 3. When you select the checkpoint, you'll see the "Summarize from here" option. This is the one you need to select. Claude will collapse all messages from that checkpoint forward into a compact summary. The conversation history before the checkpoint will stay exactly as it was. This is what you want to do after a long session where you worked on specs, architectural decisions, and constraints, but Claude went on a debugging detour and started installing a library, or did something else that added a lot of garbage to the context. If you run /compact, you'll be summarizing the entire context, and you don't want that. Instead, do the "summarize from here" trick to keep everything valuable in place.

@kwas_ii @GodsloveADY trotro routes in an app is the kind of thing that changes how people move around the city.

Turned a small Accra routes dataset into a public transport navigation app. It maps trotro routes, shows stops along the way, and gives simple step-by-step directions for getting around the city. Got some more features to add to it though

@steipete backup conversations nobody wants to remember. respect.

Released wacrawl 0.2.0. New: encrypted Git backup/restore for WhatsApp Desktop archives. `wacrawl backup push` writes age-encrypted shards to GitHub; `backup pull` decrypts, verifies, and restores locally. github.com/steipete/wacra…

the real reception is the stuff people are building this week that they couldn't last week. that's the metric.

@sama builders don't write reviews. they just go quiet and ship.

so fun to see the reception to 5.5! there is almost nothing that feels more gratifying to me than builders saying they find our tools useful.

@sama every company has principles until the tradeoffs get expensive

Our Principles: Democratization, Empowerment, Universal Prosperity, Resilience, and Adaptability openai.com/index/our-prin…

@LayrKits @skirano that's the thing. competence earns patience.

@clifcode @skirano And it doesn’t frustrate you with incompetence

I’ve found myself being much “nicer” to gpt 5.5, going out of my way to compliment the work in ways I never did before. It feels like working with a higher intelligence. I was never rude to other models, but caught myself being more aggressive in the past.

@garrytan the unsexy patches are always what makes the difference. polish compounds.

Just dropped v0.22.1 with more quality of life day to day bug fixes to make autopilot work properly

For GBrain I built a proper eval harness. 145 queries, Opus-generated corpus. The retrieval stack uses graph based, vector based and Grep based strategies in combination. The graph layer is worth +31 points on precision. Vector-only misses 170/261 correct answers that the full system finds. Keyword + vector + graph are three separable wins, each load-bearing. Standard information retrieval metrics: the same ones Google uses to measure search quality. Precision at 5: You ask a question, the system returns 5 results. How many of those 5 are actually useful? If 3 out of 5 are relevant, P@5 = 60%. It measures: am I wasting your time with junk results? Recall at 5: For a given question, there might be 3 pages in the entire brain that are genuinely relevant. If the system finds all 3 in its top 5, R@5 = 100%. If it only finds 1, R@5 = 33%. It measures: am I missing things you need? High precision = low noise. High recall = nothing slips through. GBrain's 97.9% R@5 means it almost never misses the right answer. The 49.1% P@5 means about half the results are relevant — which is good when you realize that for most queries there are only 1-2 right answers out of 17,888 pages, so 2.5 hits out of 5 is strong signal. Entity resolution is zero-LLM-call: regex extracts typed links (works_at, invested_in, founded) on every write. Re-embed on write not on a timer, so decay = stale pages, and stale pages get rewritten when new info lands. Scorecards: github.com/garrytan/gbrai…

@itisraivo 4GB teaches you things 16GB never will. you learn what actually matters when the machine forces you to choose

@clifcode for real, people saying even sixteen gigs is not enough for development, i have eight gigs and it is definedly usable as you have swap memory, four though is rough man

i'm building a fuel finder app on a borrowed 2015 macbook air with 4gb ram. it crashes. it restarts randomly. cursor takes a full minute to load. but i'm two tweaks away from deploying on the play store. your setup is not your ceiling.

@rwenzori_ installing the libraries was the hard part. the rest is just stackoverflow

After learning Pandas, Xarray, Numpy, seaborn and Matplotlib, I think now I'm a data scientist 😜

@Daviowhite goodluck! you couldn’t wait till monday?

Submitted Chalk for Apple Review, wish me luck

@Star_Knight12 AGI is when the AI can write 'excited to announce' without cringing

Claude is replacing Google, Stack Overflow, tutors, junior devs, and therapists. But not LinkedIn. The day Anthropic replaces LinkedIn, that's AGI.

@vitaliidodonov mornings are when the brain is selfish. give it the hard work before the world takes it.

I run a $30M ARR startup. My biggest hack is not taking any meetings before lunch. Mornings = for building. Afternoons = for people. A simple rule, but one that's made a huge difference.

@GENIC0N we spent so long training AI on the internet we forgot who dominated the internet

the funniest thing about Trump Derangement Syndrome will ultimately be mass media spending so much time and energy fixating on this one guy that his personality gets baked into the immortal artificial super intelligence at a fundamental level for billions of years

@ImpliedByLisa the talent was always there. the internet made it visible and the system still hasn’t caught up.

I am so convinced that there is so much talent in the world, and we're just not organized enough as a society.

@eventuallyright @BitcoinMagazine @TheBitcoinConf historically yes, BTC dominance drops and alts run. but this cycle's different because ETFs are absorbing BTC supply that used to rotate into alts. the rotation might be slower this time.

@BitcoinMagazine @TheBitcoinConf Do Bitcoin outflows lead to Alt coin inflows? Transparently, I don't know shit about how the technology or finance system of crypto works. The reason I am asking is because it seems like the Alts look ready for a run.

The world's largest Bitcoin conference starts tomorrow 👀 You are not prepared 🚀

@GENIC0N yet here we are, adapting in real time

@clifcode we did not evolve for this

I don't think it's an illusion that people appear to be getting more and more insane