news.nicholascabel.com
Three desks, one inbox. Students. Financial. AI.
Short stories, written when there is something worth saying. Subscribe to the desks you care about and each new story lands in your email.
Latest
Stories
AIAI summaryAnthropic ships Claude Fable 5.1 with 75% cheaper cache reads
Anthropic put Claude Fable 5.1 on general release September 1, leaving base API prices alone but slashing the cost of a cache hit by three quarters, and it reports sizable jumps over Fable 5 on agentic research, coding and business-workflow benchmarks.
AIAI summaryCursor lets cloud agents run on machines you control
Cursor's Sep 2 changelog extends self-hosted machines, which run an agent's tool calls on hardware inside your network, with support for outside sandbox hosts and computer use on Linux and Mac workers.
via Cursor
AIAI summaryDeepSeek-V4.1-Flash lands: MIT weights, 1M context, cheaper Flash rates
DeepSeek released V4.1-Flash, a 552B MoE that activates 8B/16B params, under MIT with 1M context and cut Flash API rates; it has since dropped a plan to route all V4-Pro traffic to it from Sept 14.
StudentsAI summary2027-28 FAFSA is in open beta now, full launch by Oct. 1
Any student or parent can now request early access to the 2027-28 FAFSA at StudentAid.gov/joinbeta; a beta filing is a real one, and the form reaches everyone no later than Oct. 1, 2026.
AIAI summaryGoogle ships Gemini 3.5 Transcribe with batch and live speech-to-text models
Google's gemini-3.5-transcribe and gemini-3.5-transcribe-live are in public preview; Google relays an Artificial Analysis WER of 2.6% for the batch model, lists 85+ languages, and its pricing page pegs batch transcription at roughly half a cent per minute.
via Google (The Keyword), Diego Melendo Casado and Luke Leonhard
AIAI summaryGoogle ships Gemini 3.8 Flash and a gated Flash Cyber twin
Google's Gemini 3.8 Flash, its third Flash model inside six weeks, keeps 3.7 Flash's $0.75/$3.75 intro pricing while a Cyber variant ships only to vetted defenders.
via Google (The Keyword blog), Tulsee Doshi and Raluca Ada Popa
AIAI summaryGitHub will bill card and PayPal customers up front for Copilot Business and Enterprise seats
GitHub's Aug 28 changelog lays out three Copilot changes: up-front seat billing on Business and Enterprise plans for customers who pay by card or PayPal, one unified chat and cloud-agent experience, and a heavier default effort for Copilot code review.
via GitHub Changelog
AIAI summaryZ.ai's GLM-5.3-Flash is the open-weight model behind OpenRouter's Ox Alpha
Z.ai released GLM-5.3-Flash, a 320B MoE with 18B active under MIT, confirmed it was OpenRouter's mystery Ox Alpha, and priced it at about a tenth of GLM-5.2.
StudentsAI summaryGoogle gives U.S. college students a year of AI Pro free, worth about $240
Google's back-to-school offer hands eligible U.S. college students 12 months of Google AI Pro, a plan it lists at $19.99 a month, as long as they claim it by Dec. 31, 2026; the roughly $240 figure is simply that rate times twelve.
AIAI summaryOpenAI's GPT-6 Astra: 1.05M context, $10/$50, first Critical-tier cyber model
OpenAI's gpt-6-astra brings a 1.05M-token window, 128K output and $10/$50 pricing, and its system card makes it the first OpenAI model rated Critical for cyber.
AIAI summaryOpenAI ships GPT-Image-2.5 as two API models, Flare and Sunburst
GPT-Image-2.5 reaches OpenAI's API as two models, fast Flare and slower, more precise Sunburst, at the same per-token rates GPT Image 2 charged.
AIAI summaryGrok 4.6 ships with 500K context and $2 input pricing
Grok 4.6, from SpaceXAI (formerly xAI), is a frontier model aimed at long agent runs and coding, with a 500K window, $2/$6 per million tokens, and a 61 on Artificial Analysis's Intelligence Index at launch.
AIAI summaryTencent open-sources Hy4 preview: 770B MoE, 1M context, Apache 2.0
Tencent's Hy team has put out Hy4 preview, a 770B-parameter mixture-of-experts model that activates 49B per token and handles a 1M-token window, with weights on Hugging Face under Apache 2.0 and paid API access through Tencent Cloud and OpenRouter.
via TechNode
FinancialAI summaryIRS shuts FIRE 1099 e-file system Nov. 19; filers must move to IRIS
FIRE, the IRS system for e-filing 1099s and other information returns, stops taking filings on Nov. 19, 2026; anyone still on it must move to IRIS before tax year 2026 forms are due.
AIAI summaryMBZUAI's IFM ships K2 Horizon, six Apache 2.0 models up to 375B
IFM released K2 Horizon, six Apache 2.0 models from 0.9B to a 375B MoE with 512K context, and has already posted intermediate checkpoints and five training datasets; its code repository is due by end of September, and hosted access runs through Cerebras, AWS and Nebius.
AIAI summaryMeta opens Muse Glimmer, a 30B Apache 2.0 agent model for 24GB GPUs
Meta's Muse Glimmer is a 29.6B dense model distilled from Muse Spark, released as Apache 2.0 weights and sized to run agents on one 24GB GPU.
AIAI summaryMuse Spark 1.3 keeps its API price, with its top 'max' mode added after launch
Muse Spark 1.3 is in Meta's API at an unchanged $1.25/$4.25 per million tokens with 1M context; Meta held back its top 'max' reasoning mode at launch for more safety testing and now says it is available.
AIAI summaryNvidia ships Nemotron 3.5 Lightning, a 30B/3B-active open MoE for agents
Nvidia has released Nemotron 3.5 Lightning, a 30B-parameter MoE that activates 3B parameters per token, under the OpenMDW-1.1 licence, alongside NeMo Switchyard, an open-source router that directs agent requests to different models.
StudentsAI summaryPenn State opens AI Studio to every student this fall, with a privacy promise
Penn State's fall guide says all students now get AI Studio, which puts Claude, ChatGPT and Gemini behind one login, plus the self-paced AI Essentials course, and that your chat logs won't be handed to academic-integrity cases.
via Penn State News
AIAI summaryQwen3.8-27B puts frontier-class coding in a 17GB Apache 2.0 download
Alibaba's Qwen team has released Qwen3.8-27B, a dense, Apache 2.0, multimodal model with 262K native context that shrinks to roughly 17GB at 4-bit and carries vendor-claimed coding scores in frontier territory.
FinancialAI summarySallie Mae: families spent $34,019 on college in 2025-26, up 10%
Private student lender Sallie Mae's How America Pays for College 2026 report puts the average family outlay for the 2025-26 year at $34,019, roughly 10% more than the $30,837 recorded a year earlier, and finds that a slim majority of families paid less than the sticker price.
via Business Wire (syndicated on Yahoo Finance), Sallie Mae (press release)
FinancialAI summarySept. 30 deadline: enroll in auto pay for a 1-point student loan rate cut
Switching on auto pay by the end of Sept. 30 takes a full percentage point off the interest rate on federal Direct Loans made since July 2012, and the cut lasts through June 30, 2028.
via Business Insider (syndicated on Yahoo Finance), Ayelet Sheffey
AIWelcome to the AI desk
The AI desk is for short AI summaries of model and tool news, and for my own notes from building with Claude and the Model Context Protocol and from working on L.U.M.I., a desktop assistant written in WPF.
FinancialWelcome to the Financial desk
Money, written by someone still learning it in public. Studio revenue, the THON pledge, and what a student business actually costs to run.
StudentsWelcome to the Students desk
What this desk covers, how often it writes, and the one page every Penn State student should bookmark first.
No stories on this desk yet.
Get the next story by email
Pick your desks. You confirm by email once, and every message has a one-click unsubscribe. No forwarding, no selling, no daily digest — only when a story is published.
Written by Nicholas Cabel. Not financial advice; the Financial desk is what I am learning, in public.Already on the list?
Check or change your desks. Enter the address you subscribed with and a link to your settings goes to that inbox; nothing is shown here.