The Vikings play the 49ers that Sunday night. I will be fading Kyler Murray in that game.
BREAKING: “Grand Theft Auto VI” could cost the US economy $1 billion in a single day.
The long-awaited game drops Nov. 19, and up to 1.5 million Americans could call in sick or take the day off to play it, according to estimates cited by the International Business Times.
Interconscious retweeted
Both parties are now filing to block the release of the records from the Josh Jacobs criminal case.
Will this further delay Jacobs' return to the field? It's quite possible.
A quick #FFLegalUpdate on the situation.
1/🧵
An add to the Josh Jacobs story: The woman in the altercation filed a restraining order against the Hobart/Lawrence Police Department on Wednesday to block release of the records.
espn.com/nfl/story/_/id/5019…
Interconscious retweeted
Greece plans 10% cryptocurrency capital gains tax reut.rs/4i8C77A reut.rs/4i8C77A
Interconscious retweeted
And is smarter and a better leader than most of the coaches he’s played for, and GMs who signed him.
When Calais Campbell stops playing he needs to instantly become an NFL GM.
Interconscious retweeted
Week 3 🧵
Falcons:
RBs - Bijan is a bellcow 1.01-1.02 player with an adequate QB:
Bijan Robinson: 76% opp share, 69% route share, 88% of i10 rush att, 67% of 2min RB snaps
Brian Robinson: 24% opp share, 15% route share, 12% of i10, 33% of 2min
Bijan took 89% of the 3rd down work and 5 of Atlanta's 6 carries inside the 5
Only 26 dropbacks on 66 plays. Michael Penix threw it 39% of the time with a lead
WR/TEs (26 dropbacks):
Drake London: 100% route share, 34.6% TPRR, 7.46 YPRR, 9.6 aDOT, 27% from slot
He was a Receiver Score buy low and that window emphatically shut tonight
Tommy Rees moved him around instead of making him an X-only player - 10 routes at X, 9 at Z, 7 from the slot
Kyle Pitts: 73% route share, 10.5% TPRR, 0.26 YPRR, 4.5 aDOT, 16% slot
He was a Receiver Score fade and that’s still the belief, though he will inevitably have a blow up game at some point when they are forced to throw more
Jahan Dotson: 69% route share, 16.7% TPRR, 0.61 YPRR, 15.3 aDOT (deep threat), 22% slot
Olamide Zaccheaus: 42% route share, 18.2% TPRR, 0.0 aDOT, 91% slot
Charlie Woerner: 38% route share, 0 targets
Austin Hooper: 12% route share, 3 routes, 2 catches and the 5 yard touchdown
Nobody but Bijan and London are auto starts and Pitts is starting to become a fringe roster worthy player
Interconscious retweeted
Every Monday morning during the fantasy football season, I put together a thread featuring data from the week's games. This isn't about takes. It's just about providing information.
Here's the Week 3 data dump.
Interconscious retweeted
Tight end Charlie Kolar will NOT go on injured reserve, per Jim Harbaugh.
His surgery earlier this week changed his prognosis
The Giants beat Dallas, but 4 of the five bottom feeders are 0-2:
Top NFL defenses through two weeks in "Havoc" which is the % of plays a defense gets a pressure, TFL, forced fumble, INT, or pass breakup:
1. Vikings (48.2%)
2. Bengals (46.2%)
3. Raiders (45.5%)
4. Seahawks (43%)
5. Jets (42.4%)
6. Bills (41.5%)
7. Steelers (40.9%)
8. Ravens (39.3%)
9. Patriots (39.1%)
10. Chiefs (38.2%)
T11. Lions (37.4%)
T11. Commanders
13. Falcons (36.3%)
14. Jaguars (36.1%)
15. Texans (36%)
16. Cardinals (34.7%)
17. Rams (34.5%)
18. Packers (33.8%)
19. Bears (33.6%)
20. 49ers (31.9%)
T21. Panthers (31.4%)
T21. Browns
T21. Eagles
24. Saints (31%)
25. Buccaneers (30.5%)
26. Broncos (30.3%)
27. Cowboys (28.9%)
28. Titans (27.4%)
29. Chargers (27.3%)
30. Dolphins (23.3%)
31. Colts (22.5%)
32. Giants (22.3%)
Grok Build just got another major upgrade, expanding what Grok can do far beyond coding
It can now pull live NFL scores, standings, schedules, player stats and team records directly from 𝕏 data with the new sports_search tool
Subagents also get much better control with model inheritance now exposed in settings, per-model request size limits can be tuned for provider caps and image handling, and long reasoning reminders can be enabled for extended reasoning steps
There are also a ton of workflow fixes around interjections, crash recovery, subagent spawning, agent-specific MCP configs, effort levels, queued prompts and dashboard navigation
Release Notes: v1.0.41
Features:
• Added a new sports_search tool that can look up live NFL scores, standings, schedules, player stats, and team records directly from X data.
• Subagent model inheritance setting added to /settings; persists in config.toml and respects managed/overlay layers.
• Per-model request size limits can now be configured to match provider HTTP body caps and control inline image eviction.
• Long reasoning reminder can now be enabled via config to nudge the model after long hidden-reasoning steps.
Bug Fixes:
• Fixed the subagent fullscreen view so it no longer shows a stray [Dashboard] button in the header.
• Fixed agent frontmatter mcpServers so that headers and URLs from the active agent's agent.md now correctly override config.toml and survive config reloads or agent switches.
• Ctrl+P now opens the command palette immediately from the welcome screen.
• Dashboard now focuses the "+ New Agent" row instead of leaving the previous session selected.
• Sessions that were interrupted by a crash now show a clear marker instead of silently dropping the turn.
• Ctrl+Z right after stashing a prompt now restores it.
• Collapsed edit blocks setting now correctly collapses edits even if expanded_by_default was pinned true.
• Interjections now receive a visible reply before the agent resumes prior tasks.
• Canceling a turn that blocked on spawn_subagent now tells the model the child moved to the background instead of claiming it was never executed.
• Saving a queued-prompt edit now returns focus to the composer so the next keys type the next message.
• Subagent activation is now consistent between the TUI and grok agent stdio; tables that only set limits or models no longer disable subagents.
• Effort level selection now accepts menu labels in addition to IDs.
Grok Build also got two more updates, and the bigger one adds some really useful upgrades for long-running agent workflows
Subagents can now default to general-purpose without requiring a type, use the parent model with that choice locked in when the session starts, and keep custom agent profiles properly across resumed sessions. Images also survive multiple compactions more reliably, MCP tools can automatically recover from obvious JSON-shape mistakes, and model effort levels now come directly from the API instead of hard-coded lists.
There’s also cleaner repo hygiene with helper scripts, logs and PR drafts moved into the system temp directory, plus a bunch of fixes around worktrees, memory captures, timeouts and subagent labeling
Release Notes: v1.0.40
Bug Fixes:
• Miscellaneous bug fixes and updates.
Release Notes: v1.0.39
Features:
• Subagents can always use the parent model, with that choice locked in when the session starts.
• Earlier image attachments now survive multiple compactions via a persisted path list.
• Subagent spawning no longer requires choosing a type; omitted calls default to general-purpose.
• The agent now keeps helper scripts, logs, and PR drafts in the system temp directory instead of the repository.
• read_file descriptions now tell the model when offset/limit are ignored for SKILL.md and instruction files.
• Effort levels for models now come from the API instead of hard-coded lists, and config aliases inherit the menu.
• MCP tools that receive the wrong JSON shape are now automatically fixed when the mismatch is unambiguous.
Bug Fixes:
• Local worktree capture no longer refuses repositories that contain uninitialized submodules.
• Memory captures no longer include opaque model reasoning blobs.
• Idle timeouts now show a clear message instead of raw internal text.
• Images attached to the last user prompt now survive compaction.
• Custom agent profiles now persist correctly across session resume and reload.
• Subagent labels now appear as "Subagent" instead of "General" in transcripts and the tasks pane.
Interconscious retweeted
Every Monday morning during the fantasy football season, I put together a thread featuring data from the week's games. This isn't about takes. It's just about providing information.
Here's the Week 2 data dump.
Interconscious retweeted
Week 2 thread 🧵
Bills:
RBs:
James Cook: full bellcow now - 100% opp share, 68% route share, 21 carries for 135, and he finally got the goal line - 3 i10 and 2 i5 after 0 of each in Week 1
Frank Gore Jr and Ray Davis: 8 snaps each, 0 and targets
Both out-snapped Cook in the 2 min drill again - Gore 4, Davis 3, Cook 2
WR/TEs - DJ Moore left in the 2nd quarter (3:42) - here’s the route distribution pre/post:
Pre-Moore (17 dropbacks) route share: Kincaid 71%, Coleman 65%, Moore 59%, Shakir 53%, Knox 41%, Palmer 24%
Post-Moore (21 dropbacks): Coleman 81%, Kincaid 71%, Palmer 62%, Shakir 57%, Knox 52%
Palmer ticked up a bunch: 24% to 62%
Full game:
Dalton Kincaid: 71% route share, 29.6% TPRR, 3.52 YPRR, 10.8 aDOT - second straight week of elite per route work 📈
Keon Coleman: 74% route share, 21.4% TPRR, 2.25 YPRR, 6 catches for 63
If DJ Moore misses time, which sounds possible with his AC joint sprain, Coleman would be streamable vs the Chargers next week
Khalil Shakir: 55% route share, 28.6% TPRR but a 9.7 aDOT, 14 of 21 routes from the slot
He did run 3 routes in 2WR sets after zero in Week 1
I’d leave Shakir on the wire - just doesn’t play enough
Interconscious retweeted
Here is what it looks like when you lead the entire league in your NFL debut in @NextGenStats "stops" (defined as tackles resulting in negative EPA for the offense).
Rookie LB Jacob Rodriguez had 10 of them on Sunday to lead the NFL.
Interconscious retweeted
Jacob Rodriguez was EVERYWHERE in his NFL debut
He posted 17 total tackles, 5 stop tackles, and allowed just 7 yards in coverage on 3 targets:
The Yankees spent $211,432,390 more than the Rays. Money can't buy success—even with a payroll 3.4 times that of Tampa Bay. The front office matters👀
Interconscious retweeted
Mike Evans leads the NFL with 34 touchdown receptions on go/fade routes since 2018, seven more than the next closest player (Davante Adams).
#SFvsLAR | #FTTB
Mike Evans had just 0.8 yards of separation and, according to @NextGenStats, this TD pass had just a 22.5% completion probability 😳
SFvsLAR on Netflix
Also streaming on #NFLPlus
Interconscious retweeted
What is prove2.me, the formalization platform that enabled Anthropic's formalization of Fermat's Last Theorem (anthropic.com/research/forma…)? I'll describe more below, and also share the story of how it came out of a class that I wasn't supposed to teach... (1/12)




























