Travis Webb

i bought GPUs instead of paying off my house early. anyone else?

@TheAIShrink @jun_song at q4 it would run on a single rtx 6000 pro

@jun_song Gemma-120b on local hardware needs a datacenter-sized GPU. It’s an oligopoly masquerading as a wishlist

The local AI models we're all waiting for, built for personal hardware sizes: > Google Gemma-120b > Nvidia Nemotron4 Ultra/Super > Qwen4-27b/122b > Deepseek-V4-Flash GA > Minimax-M3.5 > GLM-5.5-Flash

@paulmichaeldev let them keep saying it and learn the hard way. it's more fun

Can people please stop saying "software is solved". No, it isn't, and if you knew anything about software engineering and product development, you would know it isn't. Pipe down, bro.

@omkarships no, but its crazy good for how small it is

@tjwebb_ Is it better than claude opus 5

@omkarships qwen3.6. i waited too long to try it because i thought it was too small to be interesting. it's very good

Which AI model has impressed you the most recently?

@solopribuilds they thought winning the AI race meant helping people generate neat images for their ads

Seriously, how can Google still not be leading the AI race with all it's resources

@DanielSmidstrup i check that it works, but no i don't look at it

Are you still checking the code written by AI?

@sflorimm its the end of closed-source APIs. my GPU will never increase my token price

if token prices 5x, is that the end of vibe coding?

@cheaf25master closed source APIs to generate code at build time open-weights models on vllm to do things during runtime

If you're starting an AI startup today, what would you build on? - Open-source models - Closed-source APIs

@kyle_e_walker I liked this so much I made one that covers every parcel in the US: landstats.com/apps/value-sky…

850,000 New York City parcels, visualized in 3D by estimated market value per square foot, with an interactive legend filter. Powered by my R packages mapgl and freestiler.

@EightTwo_Three @petergostev i keep hearing about this compute crunch, but i've never had claude or codex refuse a request because of capacity...

@petergostev Compute crunch, no? We’re still in a research race imo. Current plans are heavily subsidized and cap usage, higher cost plans would lead to higher usage expectations. Your already getting hit with loads of api demand why add to it?

I'm still confused why OpenAI or Anthropic don't have $500, $1000, $2000 plans. Don't they want the money?

@g_bobi3827 @petergostev they do have the "call us" price. big companies are not paying sticker

@petergostev if they were midsize company they'd have "enterprise" plan or whatever with "call us" for price. but since they're fortune 500 or whatever their "enterprise" is government contracts.

@yuzu_jpg after i've run out of claude and codex quota and if my local qwen36 is busy

Do people still use Gemini?

@SpaceCoastCRE @claudeai where do you get your parcel data?

A boutique shop in Melbourne running institutional-grade infrastructure with a team of eight with @claudeai A morning brief that knows about my calls. At 7am it reads my entire team's emails, work texts, and the transcripts of calls from the past 24 hours and writes me one page. What’s on whose desk. Who's waiting on me, what I said I'd do, what's gone quiet. It carries items forward so something I skipped Tuesday is still there Thursday. Wiring in the phone system was key, because the call is where the real information is and is where things fall through the cracks. New listings scored against my buyers. A file of what each active buyer wants, size, use, price, submarket. Every morning it checks everything that hit the market in Brevard against it. Then it creates drafts to customers for me to review. Listing performance pulled automatically. It scrapes our emails, texts, calls and internal notes PLUS the platforms like Crexi and LoopNet for views, saves, and inquiries across all my listings and stacks it up. On the 1st and the 15th that becomes an owner report for every seller and landlord. Blog posts off county agendas. At 10am it reads every commission agenda for the week, local news, and EDC announcements and drafts a post on whatever's worth writing about IN MY VOICE. Social cards from my own deal data. Monthly, it pulls last month's new listings / closed deals out of our pipeline and generates the Just Sold and Just Leased / New to Market graphics with captions and posts automatically to our company social accounts. Lease follow ups. A file with every lease we've done. It finds anyone hitting month 1, month 3, or six months from expiration and drafts / prompts the check-in via email and text. A contract deadline tracker. It sees an executed contract in my inbox, finds the effective date, and calculates escrow, inspection, financing, title, survey, and closing, then drafts the timeline email to the client and all parties on the file. A monthly prospecting list. It reads county parcel data and finds owners worth a call. Ten plus year holds, out of county mailing addresses, people who own multiple parcels, recent sales nearby. Then it writes the reason to call for each one. Reddit and LinkedIn drafts. It finds threads where someone's asking a question I can actually answer and writes the comment. Prompts me to go and hit send. BPOs. It writes the full first draft of every broker price opinion in our exact format and branding, whether it's a business sale or a lease rate analysis. I correct numbers and judgment calls. OMs. Give it the deal file and it assembles the offering memorandum in our template, our branding, our language. The formatting used to take longer than the thinking. Expert witness work. It builds my presentation exhibits for testimony, same format, same branding, from my analysis. Attorney introductions. I asked for every real estate attorney in the county. It built a list of 200 and drafted a personal intro to each one asking for coffee. I deadass have 12 coffee dates next week. Developer matching. I listed a development parcel. It found 37 local developers who build that product type and wrote each one a short pitch with the site details and my bio. Books for my CPA. It read two years of bank statements and built the income and expense workbook. And then the Jobs that watch the other jobs to make sure everything debugs itself. @claudeai is nuts y'all

@giswqs need a parcel layer? landrecords.us

GeoLibre v2.3.0 is here! GeoLibre is a free and open-source, lightweight, cloud-native GIS platform for visualizing, exploring, and analyzing geospatial data. It runs everywhere you do, in the web browser, on the desktop, on mobile, and inside Jupyter notebooks, all while keeping your data local and private. This release brings a legend that writes itself from your symbology, a new GeoLens catalog browser, and 200+ GeoLibre Rust geoprocessing tools running entirely in the browser. What's new in v2.3.0 - Automatic on-map Legend: the legend builds itself from your visible layers, with class rows for graduated, categorized, rule-based, and expression styling, gradient bars for heatmaps and raster colormaps, and land-cover labels from a Raster Attribute Table. Rename, hide, reorder, or add your own entries, and it saves with the project. - Symbology swatches in the Layers panel: every row shows a dot, line, square, or image glyph in the layer's own color, so a tall layer stack reads at a glance. - GeoLens catalog browser: connect to a self-hosted GeoLens server, search its catalog, and add datasets as vector tiles, GeoJSON, or rendered raster tiles. - Emerging Hot Spot Analysis: build a space-time cube from timestamped points and classify every cell as a new, intensifying, persistent, diminishing, sporadic, oscillating, or historical hot or cold spot, all client side. - Mosaic time series: the Time Slider now steps through MosaicJSON and STAC collections of many COGs per date, on either a GPU or a WASM rendering engine. - Copy and paste layer styles: give a whole set of layers one consistent look without restyling each in turn. - Shareable tool links: deep-link any Whitebox tool with a ?tool= URL that opens the dialog preselected and pre-fills the form, with a Copy link button to build it for you. - Smarter data loading: pick which layers to load from a multi-layer GeoPackage, import CSVs whose coordinates are in any projected CRS, and read a raster's real CRS, pixel size, and extent from the metadata dialog. - Multiple AI profiles: define several provider, model, and credential setups, pick a default, and switch between them from the assistant panel. Try it out - Launch GeoLibre Web: web.geolibre.app - GitHub: github.com/opengeos/GeoLi… - Documentation: geolibre.app - Release notes: github.com/opengeos/GeoLi… #GIS #Geospatial #OpenSource #RemoteSensing #MapLibre #GeoLibre

@WescheNex1q it's even the best model for those of us with 2x H200s

After all this time.. qwen 27b is still the best model for anyone with a single spark

@teropa the perils of trying to have one model do everything

what a time to be alive

@cybergirl_io the sans serif attack

In cybersecurity What do you call this type of attack?

@Mr_Salio at least give us an updated gemma4 in the meantime. it's already outclassed by every other open-weights release

🚨 Gemini 3.5 Pro Update -Google appears to be putting increasing focus on Gemini 4 and skip Gemini 3.5 pro -As you know Gemini 3.5 Pro has been delayed multiple times -Coding performance reportedly failed to meet internal expectations -Gemini 3.6 Flash has also reportedly slipped -Gemini 4 is described as a much larger and more ambitious frontier model -Google has said coding and autonomous agents are now top priorities -stopped treating Gemini 3.5 Pro as its big comeback model and shifted most of its resources toward Gemini 4 instead. Do you think Google should keep pushing Gemini 3.5 Pro, or Skip?